Showing 1–50 of 723 results for author: Wu, B

Step-Audio 2 Technical Report

Authors: Boyong Wu, Chao Yan, Chen Hu, Cheng Yi, Chengli Feng, Fei Tian, Feiyu Shen, Gang Yu, Haoyang Zhang, Jingbei Li, Mingrui Chen, Peng Liu, Wang You, Xiangyu Tony Zhang, Xingyuan Li, Xuerui Yang, Yayue Deng, Yechang Huang, Yuxin Li, Yuxin Zhang, Zhao You, Brian Li, Changyi Wan, Hanpeng Hu, Jiangjie Zhen , et al. (84 additional authors not shown)

Abstract: This paper presents Step-Audio~2, an end-to-end multi-modal large language model designed for industry-strength audio understanding and speech conversation. By integrating a latent audio encoder and reasoning-centric reinforcement learning (RL), Step-Audio 2 achieves promising performance in automatic speech recognition (ASR) and audio understanding. To facilitate genuine end-to-end speech convers… ▽ More This paper presents Step-Audio~2, an end-to-end multi-modal large language model designed for industry-strength audio understanding and speech conversation. By integrating a latent audio encoder and reasoning-centric reinforcement learning (RL), Step-Audio 2 achieves promising performance in automatic speech recognition (ASR) and audio understanding. To facilitate genuine end-to-end speech conversation, Step-Audio 2 incorporates the generation of discrete audio tokens into language modeling, significantly enhancing its responsiveness to paralinguistic information such as speaking styles and emotions. To effectively leverage the rich textual and acoustic knowledge in real-world data, Step-Audio 2 integrates retrieval-augmented generation (RAG) and is able to call external tools such as web search to mitigate hallucination and audio search to switch timbres. Trained on millions of hours of speech and audio data, Step-Audio 2 delivers intelligence and expressiveness across diverse conversational scenarios. Evaluation results demonstrate that Step-Audio 2 achieves state-of-the-art performance on various audio understanding and conversational benchmarks compared to other open-source and commercial solutions. Please visit https://github.com/stepfun-ai/Step-Audio2 for more information. △ Less

Submitted 22 July, 2025; originally announced July 2025.

Apple Intelligence Foundation Language Models: Tech Report 2025

Authors: Hanzhi Zhou, Erik Hornberger, Pengsheng Guo, Xiyou Zhou, Saiwen Wang, Xin Wang, Yifei He, Xuankai Chang, Rene Rauch, Louis D'hauwe, John Peebles, Alec Doane, Kohen Chia, Jenna Thibodeau, Zi-Yi Dou, Yuanyang Zhang, Ruoming Pang, Reed Li, Zhifeng Chen, Jeremy Warner, Zhaoyang Xu, Sophy Lee, David Mizrahi, Ramsey Tantawi, Chris Chaney , et al. (370 additional authors not shown)

Abstract: We introduce two multilingual, multimodal foundation language models that power Apple Intelligence features across Apple devices and services: i a 3B-parameter on-device model optimized for Apple silicon through architectural innovations such as KV-cache sharing and 2-bit quantization-aware training; and ii a scalable server model built on a novel Parallel-Track Mixture-of-Experts PT-MoE transform… ▽ More We introduce two multilingual, multimodal foundation language models that power Apple Intelligence features across Apple devices and services: i a 3B-parameter on-device model optimized for Apple silicon through architectural innovations such as KV-cache sharing and 2-bit quantization-aware training; and ii a scalable server model built on a novel Parallel-Track Mixture-of-Experts PT-MoE transformer that combines track parallelism, mixture-of-experts sparse computation, and interleaved global-local attention to deliver high quality with competitive cost on Apple's Private Cloud Compute platform. Both models are trained on large-scale multilingual and multimodal datasets sourced via responsible web crawling, licensed corpora, and high-quality synthetic data, then further refined with supervised fine-tuning and reinforcement learning on a new asynchronous platform. The resulting models support several additional languages while understanding images and executing tool calls. In public benchmarks and human evaluations, both the server model and the on-device model match or surpass comparably sized open baselines. A new Swift-centric Foundation Models framework exposes guided generation, constrained tool calling, and LoRA adapter fine-tuning, allowing developers to integrate these capabilities with a few lines of code. The latest advancements in Apple Intelligence models are grounded in our Responsible AI approach with safeguards like content filtering and locale-specific evaluation, as well as our commitment to protecting our users' privacy with innovations like Private Cloud Compute. △ Less

Submitted 17 July, 2025; originally announced July 2025.

DALI-PD: Diffusion-based Synthetic Layout Heatmap Generation for ML in Physical Design

Authors: Bing-Yue Wu, Vidya A. Chhabria

Abstract: Machine learning (ML) has demonstrated significant promise in various physical design (PD) tasks. However, model generalizability remains limited by the availability of high-quality, large-scale training datasets. Creating such datasets is often computationally expensive and constrained by IP. While very few public datasets are available, they are typically static, slow to generate, and require fr… ▽ More Machine learning (ML) has demonstrated significant promise in various physical design (PD) tasks. However, model generalizability remains limited by the availability of high-quality, large-scale training datasets. Creating such datasets is often computationally expensive and constrained by IP. While very few public datasets are available, they are typically static, slow to generate, and require frequent updates. To address these limitations, we present DALI-PD, a scalable framework for generating synthetic layout heatmaps to accelerate ML in PD research. DALI-PD uses a diffusion model to generate diverse layout heatmaps via fast inference in seconds. The heatmaps include power, IR drop, congestion, macro placement, and cell density maps. Using DALI-PD, we created a dataset comprising over 20,000 layout configurations with varying macro counts and placements. These heatmaps closely resemble real layouts and improve ML accuracy on downstream ML tasks such as IR drop or congestion prediction. △ Less

Submitted 13 July, 2025; originally announced July 2025.

Comments: Under review at Asia and South Pacific Design Automation Conference (ASP-DAC'26)

Enhancing Interpretability in Software Change Management with Chain-of-Thought Reasoning

Authors: Yongqian Sun, Weihua Kuang, Chao Shen, Xidao Wen, Tinghua Zheng, Heng Liu, Shenglin Zhang, Bo Wu, Dan Pei

Abstract: In modern online services, frequent software changes introduce significant risks. To tackle this challenge, we propose SCELM (Software Change Evaluation and Lifecycle Management), an end-to-end automated framework for software change management. SCELM aims to manage software changes efficiently and precisely, significantly reducing service failures and economic losses. In modern online services, frequent software changes introduce significant risks. To tackle this challenge, we propose SCELM (Software Change Evaluation and Lifecycle Management), an end-to-end automated framework for software change management. SCELM aims to manage software changes efficiently and precisely, significantly reducing service failures and economic losses. △ Less

Submitted 12 July, 2025; originally announced July 2025.

Comments: 22 pages, 19 figures

DiffNMR: Diffusion Models for Nuclear Magnetic Resonance Spectra Elucidation

Authors: Qingsong Yang, Binglan Wu, Xuwei Liu, Bo Chen, Wei Li, Gen Long, Xin Chen, Mingjun Xiao

Abstract: Nuclear Magnetic Resonance (NMR) spectroscopy is a central characterization method for molecular structure elucidation, yet interpreting NMR spectra to deduce molecular structures remains challenging due to the complexity of spectral data and the vastness of the chemical space. In this work, we introduce DiffNMR, a novel end-to-end framework that leverages a conditional discrete diffusion model fo… ▽ More Nuclear Magnetic Resonance (NMR) spectroscopy is a central characterization method for molecular structure elucidation, yet interpreting NMR spectra to deduce molecular structures remains challenging due to the complexity of spectral data and the vastness of the chemical space. In this work, we introduce DiffNMR, a novel end-to-end framework that leverages a conditional discrete diffusion model for de novo molecular structure elucidation from NMR spectra. DiffNMR refines molecular graphs iteratively through a diffusion-based generative process, ensuring global consistency and mitigating error accumulation inherent in autoregressive methods. The framework integrates a two-stage pretraining strategy that aligns spectral and molecular representations via diffusion autoencoder (Diff-AE) and contrastive learning, the incorporation of retrieval initialization and similarity filtering during inference, and a specialized NMR encoder with radial basis function (RBF) encoding for chemical shifts, preserving continuity and chemical correlation. Experimental results demonstrate that DiffNMR achieves competitive performance for NMR-based structure elucidation, offering an efficient and robust solution for automated molecular analysis. △ Less

Submitted 9 July, 2025; originally announced July 2025.

PresentAgent: Multimodal Agent for Presentation Video Generation

Authors: Jingwei Shi, Zeyu Zhang, Biao Wu, Yanjie Liang, Meng Fang, Ling Chen, Yang Zhao

Abstract: We present PresentAgent, a multimodal agent that transforms long-form documents into narrated presentation videos. While existing approaches are limited to generating static slides or text summaries, our method advances beyond these limitations by producing fully synchronized visual and spoken content that closely mimics human-style presentations. To achieve this integration, PresentAgent employs… ▽ More We present PresentAgent, a multimodal agent that transforms long-form documents into narrated presentation videos. While existing approaches are limited to generating static slides or text summaries, our method advances beyond these limitations by producing fully synchronized visual and spoken content that closely mimics human-style presentations. To achieve this integration, PresentAgent employs a modular pipeline that systematically segments the input document, plans and renders slide-style visual frames, generates contextual spoken narration with large language models and Text-to-Speech models, and seamlessly composes the final video with precise audio-visual alignment. Given the complexity of evaluating such multimodal outputs, we introduce PresentEval, a unified assessment framework powered by Vision-Language Models that comprehensively scores videos across three critical dimensions: content fidelity, visual clarity, and audience comprehension through prompt-based evaluation. Our experimental validation on a curated dataset of 30 document-presentation pairs demonstrates that PresentAgent approaches human-level quality across all evaluation metrics. These results highlight the significant potential of controllable multimodal agents in transforming static textual materials into dynamic, effective, and accessible presentation formats. Code will be available at https://github.com/AIGeeksGroup/PresentAgent. △ Less

Submitted 5 July, 2025; originally announced July 2025.

Enhancing Vehicular Platooning with Wireless Federated Learning: A Resource-Aware Control Framework

Authors: Beining Wu, Jun Huang, Qiang Duan, Liang Dong, Zhipeng Cai

Abstract: This paper aims to enhance the performance of Vehicular Platooning (VP) systems integrated with Wireless Federated Learning (WFL). In highly dynamic environments, vehicular platoons experience frequent communication changes and resource constraints, which significantly affect information exchange and learning model synchronization. To address these challenges, we first formulate WFL in VP as a joi… ▽ More This paper aims to enhance the performance of Vehicular Platooning (VP) systems integrated with Wireless Federated Learning (WFL). In highly dynamic environments, vehicular platoons experience frequent communication changes and resource constraints, which significantly affect information exchange and learning model synchronization. To address these challenges, we first formulate WFL in VP as a joint optimization problem that simultaneously considers Age of Information (AoI) and Federated Learning Model Drift (FLMD) to ensure timely and accurate control. Through theoretical analysis, we examine the impact of FLMD on convergence performance and develop a two-stage Resource-Aware Control framework (RACE). The first stage employs a Lagrangian dual decomposition method for resource configuration, while the second stage implements a multi-agent deep reinforcement learning approach for vehicle selection. The approach integrates Multi-Head Self-Attention and Long Short-Term Memory networks to capture spatiotemporal correlations in communication states. Experimental results demonstrate that, compared to baseline methods, the proposed framework improves AoI optimization by up to 45%, accelerates learning convergence, and adapts more effectively to dynamic VP environments on the AI4MARS dataset. △ Less

Submitted 1 July, 2025; originally announced July 2025.

Comments: Under review at IEEE Transactions on Networking

Boosting LLM's Molecular Structure Elucidation with Knowledge Enhanced Tree Search Reasoning

Authors: Xiang Zhuang, Bin Wu, Jiyu Cui, Kehua Feng, Xiaotong Li, Huabin Xing, Keyan Ding, Qiang Zhang, Huajun Chen

Abstract: Molecular structure elucidation involves deducing a molecule's structure from various types of spectral data, which is crucial in chemical experimental analysis. While large language models (LLMs) have shown remarkable proficiency in analyzing and reasoning through complex tasks, they still encounter substantial challenges in molecular structure elucidation. We identify that these challenges large… ▽ More Molecular structure elucidation involves deducing a molecule's structure from various types of spectral data, which is crucial in chemical experimental analysis. While large language models (LLMs) have shown remarkable proficiency in analyzing and reasoning through complex tasks, they still encounter substantial challenges in molecular structure elucidation. We identify that these challenges largely stem from LLMs' limited grasp of specialized chemical knowledge. In this work, we introduce a Knowledge-enhanced reasoning framework for Molecular Structure Elucidation (K-MSE), leveraging Monte Carlo Tree Search for test-time scaling as a plugin. Specifically, we construct an external molecular substructure knowledge base to extend the LLMs' coverage of the chemical structure space. Furthermore, we design a specialized molecule-spectrum scorer to act as a reward model for the reasoning process, addressing the issue of inaccurate solution evaluation in LLMs. Experimental results show that our approach significantly boosts performance, particularly gaining more than 20% improvement on both GPT-4o-mini and GPT-4o. Our code is available at https://github.com/HICAI-ZJU/K-MSE. △ Less

Submitted 28 June, 2025; originally announced June 2025.

Comments: ACL 2025 Main

Towards Explainable Bilingual Multimodal Misinformation Detection and Localization

Authors: Yiwei He, Xiangtai Li, Zhenglin Huang, Yi Dong, Hao Fei, Jiangning Zhang, Baoyuan Wu, Guangliang Cheng

Abstract: The increasing realism of multimodal content has made misinformation more subtle and harder to detect, especially in news media where images are frequently paired with bilingual (e.g., Chinese-English) subtitles. Such content often includes localized image edits and cross-lingual inconsistencies that jointly distort meaning while remaining superficially plausible. We introduce BiMi, a bilingual mu… ▽ More The increasing realism of multimodal content has made misinformation more subtle and harder to detect, especially in news media where images are frequently paired with bilingual (e.g., Chinese-English) subtitles. Such content often includes localized image edits and cross-lingual inconsistencies that jointly distort meaning while remaining superficially plausible. We introduce BiMi, a bilingual multimodal framework that jointly performs region-level localization, cross-modal and cross-lingual consistency detection, and natural language explanation for misinformation analysis. To support generalization, BiMi integrates an online retrieval module that supplements model reasoning with up-to-date external context. We further release BiMiBench, a large-scale and comprehensive benchmark constructed by systematically editing real news images and subtitles, comprising 104,000 samples with realistic manipulations across visual and linguistic modalities. To enhance interpretability, we apply Group Relative Policy Optimization (GRPO) to improve explanation quality, marking the first use of GRPO in this domain. Extensive experiments demonstrate that BiMi outperforms strong baselines by up to +8.9 in classification accuracy, +15.9 in localization accuracy, and +2.5 in explanation BERTScore, advancing state-of-the-art performance in realistic, multilingual misinformation detection. Code, models, and datasets will be released. △ Less

Submitted 28 June, 2025; originally announced June 2025.

Progressive Alignment Degradation Learning for Pansharpening

Authors: Enzhe Zhao, Zhichang Guo, Yao Li, Fanghui Song, Boying Wu

Abstract: Deep learning-based pansharpening has been shown to effectively generate high-resolution multispectral (HRMS) images. To create supervised ground-truth HRMS images, synthetic data generated using the Wald protocol is commonly employed. This protocol assumes that networks trained on artificial low-resolution data will perform equally well on high-resolution data. However, well-trained models typica… ▽ More Deep learning-based pansharpening has been shown to effectively generate high-resolution multispectral (HRMS) images. To create supervised ground-truth HRMS images, synthetic data generated using the Wald protocol is commonly employed. This protocol assumes that networks trained on artificial low-resolution data will perform equally well on high-resolution data. However, well-trained models typically exhibit a trade-off in performance between reduced-resolution and full-resolution datasets. In this paper, we delve into the Wald protocol and find that its inaccurate approximation of real-world degradation patterns limits the generalization of deep pansharpening models. To address this issue, we propose the Progressive Alignment Degradation Module (PADM), which uses mutual iteration between two sub-networks, PAlignNet and PDegradeNet, to adaptively learn accurate degradation processes without relying on predefined operators. Building on this, we introduce HFreqdiff, which embeds high-frequency details into a diffusion framework and incorporates CFB and BACM modules for frequency-selective detail extraction and precise reverse process learning. These innovations enable effective integration of high-resolution panchromatic and multispectral images, significantly enhancing spatial sharpness and quality. Experiments and ablation studies demonstrate the proposed method's superior performance compared to state-of-the-art techniques. △ Less

Submitted 25 June, 2025; originally announced June 2025.

Comments: 13 pages, 9 figures

SAFEx: Analyzing Vulnerabilities of MoE-Based LLMs via Stable Safety-critical Expert Identification

Authors: Zhenglin Lai, Mengyao Liao, Dong Xu, Zebin Zhao, Zhihang Yuan, Chao Fan, Jianqiang Li, Bingzhe Wu

Abstract: Large language models based on Mixture-of-Experts have achieved substantial gains in efficiency and scalability, yet their architectural uniqueness introduces underexplored safety alignment challenges. Existing safety alignment strategies, predominantly designed for dense models, are ill-suited to address MoE-specific vulnerabilities. In this work, we formalize and systematically study MoE model's… ▽ More Large language models based on Mixture-of-Experts have achieved substantial gains in efficiency and scalability, yet their architectural uniqueness introduces underexplored safety alignment challenges. Existing safety alignment strategies, predominantly designed for dense models, are ill-suited to address MoE-specific vulnerabilities. In this work, we formalize and systematically study MoE model's positional vulnerability - the phenomenon where safety-aligned behaviors rely on specific expert modules, revealing critical risks inherent to MoE architectures. To this end, we present SAFEx, an analytical framework that robustly identifies, characterizes, and validates the safety-critical experts using a novel Stability-based Expert Selection (SES) algorithm. Notably, our approach enables the explicit decomposition of safety-critical experts into distinct functional groups, including those responsible for harmful content detection and those controlling safe response generation. Extensive experiments on mainstream MoE models, such as the recently released Qwen3-MoE, demonstrated that their intrinsic safety mechanisms heavily rely on a small subset of positional experts. Disabling these experts significantly compromised the models' ability to refuse harmful requests. For Qwen3-MoE with 6144 experts (in the FNN layer), we find that disabling as few as 12 identified safety-critical experts can cause the refusal rate to drop by 22%, demonstrating the disproportionate impact of a small set of experts on overall model safety. △ Less

Submitted 20 June, 2025; originally announced June 2025.

Comments: 9 pages, 7 figures

Dense SAE Latents Are Features, Not Bugs

Authors: Xiaoqing Sun, Alessandro Stolfo, Joshua Engels, Ben Wu, Senthooran Rajamanoharan, Mrinmaya Sachan, Max Tegmark

Abstract: Sparse autoencoders (SAEs) are designed to extract interpretable features from language models by enforcing a sparsity constraint. Ideally, training an SAE would yield latents that are both sparse and semantically meaningful. However, many SAE latents activate frequently (i.e., are \emph{dense}), raising concerns that they may be undesirable artifacts of the training procedure. In this work, we sy… ▽ More Sparse autoencoders (SAEs) are designed to extract interpretable features from language models by enforcing a sparsity constraint. Ideally, training an SAE would yield latents that are both sparse and semantically meaningful. However, many SAE latents activate frequently (i.e., are \emph{dense}), raising concerns that they may be undesirable artifacts of the training procedure. In this work, we systematically investigate the geometry, function, and origin of dense latents and show that they are not only persistent but often reflect meaningful model representations. We first demonstrate that dense latents tend to form antipodal pairs that reconstruct specific directions in the residual stream, and that ablating their subspace suppresses the emergence of new dense features in retrained SAEs -- suggesting that high density features are an intrinsic property of the residual space. We then introduce a taxonomy of dense latents, identifying classes tied to position tracking, context binding, entropy regulation, letter-specific output signals, part-of-speech, and principal component reconstruction. Finally, we analyze how these features evolve across layers, revealing a shift from structural features in early layers, to semantic features in mid layers, and finally to output-oriented signals in the last layers of the model. Our findings indicate that dense latents serve functional roles in language model computation and should not be dismissed as training noise. △ Less

Submitted 18 June, 2025; originally announced June 2025.

Breaking Thought Patterns: A Multi-Dimensional Reasoning Framework for LLMs

Authors: Xintong Tang, Meiru Zhang, Shang Xiao, Junzhao Jin, Zihan Zhao, Liwei Li, Yang Zheng, Bangyi Wu

Abstract: Large language models (LLMs) are often constrained by rigid reasoning processes, limiting their ability to generate creative and diverse responses. To address this, a novel framework called LADDER is proposed, combining Chain-of-Thought (CoT) reasoning, Mixture of Experts (MoE) models, and multi-dimensional up/down-sampling strategies which breaks the limitations of traditional LLMs. First, CoT re… ▽ More Large language models (LLMs) are often constrained by rigid reasoning processes, limiting their ability to generate creative and diverse responses. To address this, a novel framework called LADDER is proposed, combining Chain-of-Thought (CoT) reasoning, Mixture of Experts (MoE) models, and multi-dimensional up/down-sampling strategies which breaks the limitations of traditional LLMs. First, CoT reasoning guides the model through multi-step logical reasoning, expanding the semantic space and breaking the rigidity of thought. Next, MoE distributes the reasoning tasks across multiple expert modules, each focusing on specific sub-tasks. Finally, dimensionality reduction maps the reasoning outputs back to a lower-dimensional semantic space, yielding more precise and creative responses. Extensive experiments across multiple tasks demonstrate that LADDER significantly improves task completion, creativity, and fluency, generating innovative and coherent responses that outperform traditional models. Ablation studies reveal the critical roles of CoT and MoE in enhancing reasoning abilities and creative output. This work contributes to the development of more flexible and creative LLMs, capable of addressing complex and novel tasks. △ Less

Submitted 16 June, 2025; originally announced June 2025.

VGR: Visual Grounded Reasoning

Authors: Jiacong Wang, Zijian Kang, Haochen Wang, Haiyong Jiang, Jiawen Li, Bohong Wu, Ya Wang, Jiao Ran, Xiao Liang, Chao Feng, Jun Xiao

Abstract: In the field of multimodal chain-of-thought (CoT) reasoning, existing approaches predominantly rely on reasoning on pure language space, which inherently suffers from language bias and is largely confined to math or science domains. This narrow focus limits their ability to handle complex visual reasoning tasks that demand comprehensive understanding of image details. To address these limitations,… ▽ More In the field of multimodal chain-of-thought (CoT) reasoning, existing approaches predominantly rely on reasoning on pure language space, which inherently suffers from language bias and is largely confined to math or science domains. This narrow focus limits their ability to handle complex visual reasoning tasks that demand comprehensive understanding of image details. To address these limitations, this paper introduces VGR, a novel reasoning multimodal large language model (MLLM) with enhanced fine-grained visual perception capabilities. Unlike traditional MLLMs that answer the question or reasoning solely on the language space, our VGR first detects relevant regions that may help to solve problems, and then provides precise answers based on replayed image regions. To achieve this, we conduct a large-scale SFT dataset called VGR -SFT that contains reasoning data with mixed vision grounding and language deduction. The inference pipeline of VGR allows the model to choose bounding boxes for visual reference and a replay stage is introduced to integrates the corresponding regions into the reasoning process, enhancing multimodel comprehension. Experiments on the LLaVA-NeXT-7B baseline show that VGR achieves superior performance on multi-modal benchmarks requiring comprehensive image detail understanding. Compared to the baseline, VGR uses only 30\% of the image token count while delivering scores of +4.1 on MMStar, +7.1 on AI2D, and a +12.9 improvement on ChartQA. △ Less

Submitted 16 June, 2025; v1 submitted 13 June, 2025; originally announced June 2025.

Comments: 9 pages, 4 figures

TransXSSM: A Hybrid Transformer State Space Model with Unified Rotary Position Embedding

Authors: Bingheng Wu, Jingze Shi, Yifan Wu, Nan Tang, Yuyu Luo

Abstract: Transformers exhibit proficiency in capturing long-range dependencies, whereas State Space Models (SSMs) facilitate linear-time sequence modeling. Notwithstanding their synergistic potential, the integration of these architectures presents a significant challenge, primarily attributable to a fundamental incongr inuity their respective positional encoding mechanisms: Transformers rely on explicit R… ▽ More Transformers exhibit proficiency in capturing long-range dependencies, whereas State Space Models (SSMs) facilitate linear-time sequence modeling. Notwithstanding their synergistic potential, the integration of these architectures presents a significant challenge, primarily attributable to a fundamental incongr inuity their respective positional encoding mechanisms: Transformers rely on explicit Rotary Position Embeddings (RoPE), while SSMs leverage implicit positional representations via convolutions. This divergence often precipitates discontinuities and suboptimal performance.To address this impediment, we propose a unified rotary position embedding (Unified RoPE) methodology, thereby establishing a consistent positional encoding framework for both self-attention and state-space components. Using this Unified RoPE, we introduce TransXSSM, a hybrid architecture that coherently integrates the Transformer and SSM layers under this unified positional encoding scheme. At a 4 sequenceK length, TransXSSM exhibits training and inference speeds that are 42.3% and 29.5% faster, respectively, relative to standard Transformer models. It also delivers higher accuracy: under comparable settings, it surpasses a Transformer baseline by over 4% on language modeling benchmarks.TransXSSM furthermore scales more effectively: TransXSSM-1.3B gains 7.22% in average accuracy over its 320M version (versus about 6% gains for equivalent Transformers or SSMs). Our results show that unified positional encoding resolves positional incompatibility in hybrid models, enabling efficient, high-performance long-context modeling. △ Less

Submitted 18 June, 2025; v1 submitted 11 June, 2025; originally announced June 2025.

Terabyte-Scale Analytics in the Blink of an Eye

Authors: Bowen Wu, Wei Cui, Carlo Curino, Matteo Interlandi, Rathijit Sen

Abstract: For the past two decades, the DB community has devoted substantial research to take advantage of cheap clusters of machines for distributed data analytics -- we believe that we are at the beginning of a paradigm shift. The scaling laws and popularity of AI models lead to the deployment of incredibly powerful GPU clusters in commercial data centers. Compared to CPU-only solutions, these clusters de… ▽ More For the past two decades, the DB community has devoted substantial research to take advantage of cheap clusters of machines for distributed data analytics -- we believe that we are at the beginning of a paradigm shift. The scaling laws and popularity of AI models lead to the deployment of incredibly powerful GPU clusters in commercial data centers. Compared to CPU-only solutions, these clusters deliver impressive improvements in per-node compute, memory bandwidth, and inter-node interconnect performance. In this paper, we study the problem of scaling analytical SQL queries on distributed clusters of GPUs, with the stated goal of establishing an upper bound on the likely performance gains. To do so, we build a prototype designed to maximize performance by leveraging ML/HPC best practices, such as group communication primitives for cross-device data movements. This allows us to conduct thorough performance experimentation to point our community towards a massive performance opportunity of at least 60$\times$. To make these gains more relatable, before you can blink twice, our system can run all 22 queries of TPC-H at a 1TB scale factor! △ Less

Submitted 10 June, 2025; originally announced June 2025.

Step-Audio-AQAA: a Fully End-to-End Expressive Large Audio Language Model

Authors: Ailin Huang, Bingxin Li, Bruce Wang, Boyong Wu, Chao Yan, Chengli Feng, Heng Wang, Hongyu Zhou, Hongyuan Wang, Jingbei Li, Jianjian Sun, Joanna Wang, Mingrui Chen, Peng Liu, Ruihang Miao, Shilei Jiang, Tian Fei, Wang You, Xi Chen, Xuerui Yang, Yechang Huang, Yuxiang Zhang, Zheng Ge, Zheng Gong, Zhewei Huang , et al. (51 additional authors not shown)

Abstract: Large Audio-Language Models (LALMs) have significantly advanced intelligent human-computer interaction, yet their reliance on text-based outputs limits their ability to generate natural speech responses directly, hindering seamless audio interactions. To address this, we introduce Step-Audio-AQAA, a fully end-to-end LALM designed for Audio Query-Audio Answer (AQAA) tasks. The model integrates a du… ▽ More Large Audio-Language Models (LALMs) have significantly advanced intelligent human-computer interaction, yet their reliance on text-based outputs limits their ability to generate natural speech responses directly, hindering seamless audio interactions. To address this, we introduce Step-Audio-AQAA, a fully end-to-end LALM designed for Audio Query-Audio Answer (AQAA) tasks. The model integrates a dual-codebook audio tokenizer for linguistic and semantic feature extraction, a 130-billion-parameter backbone LLM and a neural vocoder for high-fidelity speech synthesis. Our post-training approach employs interleaved token-output of text and audio to enhance semantic coherence and combines Direct Preference Optimization (DPO) with model merge to improve performance. Evaluations on the StepEval-Audio-360 benchmark demonstrate that Step-Audio-AQAA excels especially in speech control, outperforming the state-of-art LALMs in key areas. This work contributes a promising solution for end-to-end LALMs and highlights the critical role of token-based vocoder in enhancing overall performance for AQAA tasks. △ Less

Submitted 13 June, 2025; v1 submitted 10 June, 2025; originally announced June 2025.

Comments: 12 pages, 3 figures

Initial Model Incorporation for Deep Learning FWI: Pretraining or Denormalization?

Authors: Ruihua Chen, Bangyu Wu, Meng Li, Kai Yang

Abstract: Subsurface property neural network reparameterized full waveform inversion (FWI) has emerged as an effective unsupervised learning framework, which can invert stably with an inaccurate starting model. It updates the trainable neural network parameters instead of fine-tuning on the subsurface model directly. There are primarily two ways to embed the prior knowledge of the initial model into neural… ▽ More Subsurface property neural network reparameterized full waveform inversion (FWI) has emerged as an effective unsupervised learning framework, which can invert stably with an inaccurate starting model. It updates the trainable neural network parameters instead of fine-tuning on the subsurface model directly. There are primarily two ways to embed the prior knowledge of the initial model into neural networks, that is, pretraining and denormalization. Pretraining first regulates the neural networks' parameters by fitting the initial velocity model; Denormalization directly adds the outputs of the network into the initial models without pretraining. In this letter, we systematically investigate the influence of the two ways of initial model incorporation for the neural network reparameterized FWI. We demonstrate that pretraining requires inverting the model perturbation based on a constant velocity value (mean) with a two-stage implementation. It leads to a complex workflow and inconsistency of objective functions in the two-stage process, causing the network parameters to become inactive and lose plasticity. Experimental results demonstrate that denormalization can simplify workflows, accelerate convergence, and enhance inversion accuracy compared with pretraining. △ Less

Submitted 5 June, 2025; originally announced June 2025.

Scaling Transformers for Discriminative Recommendation via Generative Pretraining

Authors: Chunqi Wang, Bingchao Wu, Zheng Chen, Lei Shen, Bing Wang, Xiaoyi Zeng

Abstract: Discriminative recommendation tasks, such as CTR (click-through rate) and CVR (conversion rate) prediction, play critical roles in the ranking stage of large-scale industrial recommender systems. However, training a discriminative model encounters a significant overfitting issue induced by data sparsity. Moreover, this overfitting issue worsens with larger models, causing them to underperform smal… ▽ More Discriminative recommendation tasks, such as CTR (click-through rate) and CVR (conversion rate) prediction, play critical roles in the ranking stage of large-scale industrial recommender systems. However, training a discriminative model encounters a significant overfitting issue induced by data sparsity. Moreover, this overfitting issue worsens with larger models, causing them to underperform smaller ones. To address the overfitting issue and enhance model scalability, we propose a framework named GPSD (\textbf{G}enerative \textbf{P}retraining for \textbf{S}calable \textbf{D}iscriminative Recommendation), drawing inspiration from generative training, which exhibits no evident signs of overfitting. GPSD leverages the parameters learned from a pretrained generative model to initialize a discriminative model, and subsequently applies a sparse parameter freezing strategy. Extensive experiments conducted on both industrial-scale and publicly available datasets demonstrate the superior performance of GPSD. Moreover, it delivers remarkable improvements in online A/B tests. GPSD offers two primary advantages: 1) it substantially narrows the generalization gap in model training, resulting in better test performance; and 2) it leverages the scalability of Transformers, delivering consistent performance gains as models are scaled up. Specifically, we observe consistent performance improvements as the model dense parameters scale from 13K to 0.3B, closely adhering to power laws. These findings pave the way for unifying the architectures of recommendation models and language models, enabling the direct application of techniques well-established in large language models to recommendation models. △ Less

Submitted 4 June, 2025; originally announced June 2025.

Comments: KDD'25

Infinity Parser: Layout Aware Reinforcement Learning for Scanned Document Parsing

Authors: Baode Wang, Biao Wu, Weizhen Li, Meng Fang, Yanjie Liang, Zuming Huang, Haozhe Wang, Jun Huang, Ling Chen, Wei Chu, Yuan Qi

Abstract: Automated parsing of scanned documents into richly structured, machine-readable formats remains a critical bottleneck in Document AI, as traditional multi-stage pipelines suffer from error propagation and limited adaptability to diverse layouts. We introduce layoutRL, an end-to-end reinforcement learning framework that trains models to be explicitly layout-aware by optimizing a composite reward of… ▽ More Automated parsing of scanned documents into richly structured, machine-readable formats remains a critical bottleneck in Document AI, as traditional multi-stage pipelines suffer from error propagation and limited adaptability to diverse layouts. We introduce layoutRL, an end-to-end reinforcement learning framework that trains models to be explicitly layout-aware by optimizing a composite reward of normalized edit distance, paragraph count accuracy, and reading order preservation. Leveraging our newly released dataset, Infinity-Doc-55K, which combines 55K high-fidelity synthetic scanned document parsing data with expert-filtered real-world documents, we instantiate layoutRL in a vision-language-model-based parser called Infinity-Parser. Evaluated on English and Chinese benchmarks for OCR, table and formula extraction, and reading order detection, Infinity-Parser achieves new state-of-the-art performance in both accuracy and structural fidelity, outpacing specialist pipelines and general-purpose vision-language models. We will publicly release our code and dataset to accelerate progress in robust document understanding. △ Less

Submitted 1 June, 2025; originally announced June 2025.

Comments: 16 pages, 12 figures

Report number: INF-CS-TR-2025-02

NTPP: Generative Speech Language Modeling for Dual-Channel Spoken Dialogue via Next-Token-Pair Prediction

Authors: Qichao Wang, Ziqiao Meng, Wenqian Cui, Yifei Zhang, Pengcheng Wu, Bingzhe Wu, Irwin King, Liang Chen, Peilin Zhao

Abstract: Inspired by the impressive capabilities of GPT-4o, there is growing interest in enabling speech language models (SLMs) to engage in natural, fluid spoken interactions with humans. Recent advancements have led to the development of several SLMs that demonstrate promising results in this area. However, current approaches have yet to fully exploit dual-channel speech data, which inherently captures t… ▽ More Inspired by the impressive capabilities of GPT-4o, there is growing interest in enabling speech language models (SLMs) to engage in natural, fluid spoken interactions with humans. Recent advancements have led to the development of several SLMs that demonstrate promising results in this area. However, current approaches have yet to fully exploit dual-channel speech data, which inherently captures the structure and dynamics of human conversation. In this work, we systematically explore the use of dual-channel speech data in the context of modern large language models, and introduce a novel generative modeling paradigm, Next-Token-Pair Prediction (NTPP), to enable speaker-independent dual-channel spoken dialogue learning using decoder-only architectures for the first time. We evaluate our approach on standard benchmarks, and empirical results show that our proposed method, NTPP, significantly improves the conversational abilities of SLMs in terms of turn-taking prediction, response coherence, and naturalness. Moreover, compared to existing methods, NTPP achieves substantially lower inference latency, highlighting its practical efficiency for real-time applications. △ Less

Submitted 11 June, 2025; v1 submitted 1 June, 2025; originally announced June 2025.

Comments: Accepted by ICML 2025

Addressing the Collaboration Dilemma in Low-Data Federated Learning via Transient Sparsity

Authors: Qiao Xiao, Boqian Wu, Andrey Poddubnyy, Elena Mocanu, Phuong H. Nguyen, Mykola Pechenizkiy, Decebal Constantin Mocanu

Abstract: Federated learning (FL) enables collaborative model training across decentralized clients while preserving data privacy, leveraging aggregated updates to build robust global models. However, this training paradigm faces significant challenges due to data heterogeneity and limited local datasets, which often impede effective collaboration. In such scenarios, we identify the Layer-wise Inertia Pheno… ▽ More Federated learning (FL) enables collaborative model training across decentralized clients while preserving data privacy, leveraging aggregated updates to build robust global models. However, this training paradigm faces significant challenges due to data heterogeneity and limited local datasets, which often impede effective collaboration. In such scenarios, we identify the Layer-wise Inertia Phenomenon in FL, wherein the middle layers of global model undergo minimal updates after early communication rounds, ultimately limiting the effectiveness of global aggregation. We demonstrate the presence of this phenomenon across a wide range of federated settings, spanning diverse datasets and architectures. To address this issue, we propose LIPS (Layer-wise Inertia Phenomenon with Sparsity), a simple yet effective method that periodically introduces transient sparsity to stimulate meaningful updates and empower global aggregation. Experiments demonstrate that LIPS effectively mitigates layer-wise inertia, enhances aggregation effectiveness, and improves overall performance in various FL scenarios. This work not only deepens the understanding of layer-wise learning dynamics in FL but also paves the way for more effective collaboration strategies in resource-constrained environments. Our code is publicly available at: https://github.com/QiaoXiao7282/LIPS. △ Less

Submitted 1 June, 2025; originally announced June 2025.

Leave it to the Specialist: Repair Sparse LLMs with Sparse Fine-Tuning via Sparsity Evolution

Authors: Qiao Xiao, Alan Ansell, Boqian Wu, Lu Yin, Mykola Pechenizkiy, Shiwei Liu, Decebal Constantin Mocanu

Abstract: Large language models (LLMs) have achieved remarkable success across various tasks but face deployment challenges due to their massive computational demands. While post-training pruning methods like SparseGPT and Wanda can effectively reduce the model size, but struggle to maintain model performance at high sparsity levels, limiting their utility for downstream tasks. Existing fine-tuning methods,… ▽ More Large language models (LLMs) have achieved remarkable success across various tasks but face deployment challenges due to their massive computational demands. While post-training pruning methods like SparseGPT and Wanda can effectively reduce the model size, but struggle to maintain model performance at high sparsity levels, limiting their utility for downstream tasks. Existing fine-tuning methods, such as full fine-tuning and LoRA, fail to preserve sparsity as they require updating the whole dense metrics, not well-suited for sparse LLMs. In this paper, we propose Sparsity Evolution Fine-Tuning (SEFT), a novel method designed specifically for sparse LLMs. SEFT dynamically evolves the sparse topology of pruned models during fine-tuning, while preserving the overall sparsity throughout the process. The strengths of SEFT lie in its ability to perform task-specific adaptation through a weight drop-and-grow strategy, enabling the pruned model to self-adapt its sparse connectivity pattern based on the target dataset. Furthermore, a sensitivity-driven pruning criterion is employed to ensure that the desired sparsity level is consistently maintained throughout fine-tuning. Our experiments on various LLMs, including LLaMA families, DeepSeek, and Mistral, across a diverse set of benchmarks demonstrate that SEFT achieves stronger performance while offering superior memory and time efficiency compared to existing baselines. Our code is publicly available at: https://github.com/QiaoXiao7282/SEFT. △ Less

Submitted 29 May, 2025; originally announced May 2025.

LlamaRL: A Distributed Asynchronous Reinforcement Learning Framework for Efficient Large-scale LLM Training

Authors: Bo Wu, Sid Wang, Yunhao Tang, Jia Ding, Eryk Helenowski, Liang Tan, Tengyu Xu, Tushar Gowda, Zhengxing Chen, Chen Zhu, Xiaocheng Tang, Yundi Qian, Beibei Zhu, Rui Hou

Abstract: Reinforcement Learning (RL) has become the most effective post-training approach for improving the capabilities of Large Language Models (LLMs). In practice, because of the high demands on latency and memory, it is particularly challenging to develop an efficient RL framework that reliably manages policy models with hundreds to thousands of billions of parameters. In this paper, we present Llama… ▽ More Reinforcement Learning (RL) has become the most effective post-training approach for improving the capabilities of Large Language Models (LLMs). In practice, because of the high demands on latency and memory, it is particularly challenging to develop an efficient RL framework that reliably manages policy models with hundreds to thousands of billions of parameters. In this paper, we present LlamaRL, a fully distributed, asynchronous RL framework optimized for efficient training of large-scale LLMs with various model sizes (8B, 70B, and 405B parameters) on GPU clusters ranging from a handful to thousands of devices. LlamaRL introduces a streamlined, single-controller architecture built entirely on native PyTorch, enabling modularity, ease of use, and seamless scalability to thousands of GPUs. We also provide a theoretical analysis of LlamaRL's efficiency, including a formal proof that its asynchronous design leads to strict RL speed-up. Empirically during the Llama 3 post-training, by leveraging best practices such as colocated model offloading, asynchronous off-policy training, and distributed direct memory access for weight synchronization, LlamaRL achieves significant efficiency gains -- up to 10.7x speed-up compared to DeepSpeed-Chat-like systems on a 405B-parameter policy model. Furthermore, the efficiency advantage continues to grow with increasing model scale, demonstrating the framework's suitability for future large-scale RL training. △ Less

Submitted 1 June, 2025; v1 submitted 29 May, 2025; originally announced May 2025.

Sparsification and Reconstruction from the Perspective of Representation Geometry

Authors: Wenjie Sun, Bingzhe Wu, Zhile Yang, Chengke Wu

Abstract: Sparse Autoencoders (SAEs) have emerged as a predominant tool in mechanistic interpretability, aiming to identify interpretable monosemantic features. However, how does sparse encoding organize the representations of activation vector from language models? What is the relationship between this organizational paradigm and feature disentanglement as well as reconstruction performance? To address the… ▽ More Sparse Autoencoders (SAEs) have emerged as a predominant tool in mechanistic interpretability, aiming to identify interpretable monosemantic features. However, how does sparse encoding organize the representations of activation vector from language models? What is the relationship between this organizational paradigm and feature disentanglement as well as reconstruction performance? To address these questions, we propose the SAEMA, which validates the stratified structure of the representation by observing the variability of the rank of the symmetric semipositive definite (SSPD) matrix corresponding to the modal tensor unfolded along the latent tensor with the level of noise added to the residual stream. To systematically investigate how sparse encoding alters representational structures, we define local and global representations, demonstrating that they amplify inter-feature distinctions by merging similar semantic features and introducing additional dimensionality. Furthermore, we intervene the global representation from an optimization perspective, proving a significant causal relationship between their separability and the reconstruction performance. This study explains the principles of sparsity from the perspective of representational geometry and demonstrates the impact of changes in representational structure on reconstruction performance. Particularly emphasizes the necessity of understanding representations and incorporating representational constraints, providing empirical references for developing new interpretable tools and improving SAEs. The code is available at \hyperlink{https://github.com/wenjie1835/SAERepGeo}{https://github.com/wenjie1835/SAERepGeo}. △ Less

Submitted 28 May, 2025; originally announced May 2025.

Comments: 24 pages, 5 figures

MSC Class: 22-08 ACM Class: I.2.4; I.2.7

Any-to-Bokeh: One-Step Video Bokeh via Multi-Plane Image Guided Diffusion

Authors: Yang Yang, Siming Zheng, Jinwei Chen, Boxi Wu, Xiaofei He, Deng Cai, Bo Li, Peng-Tao Jiang

Abstract: Recent advances in diffusion based editing models have enabled realistic camera simulation and image-based bokeh, but video bokeh remains largely unexplored. Existing video editing models cannot explicitly control focus planes or adjust bokeh intensity, limiting their applicability for controllable optical effects. Moreover, naively extending image-based bokeh methods to video often results in tem… ▽ More Recent advances in diffusion based editing models have enabled realistic camera simulation and image-based bokeh, but video bokeh remains largely unexplored. Existing video editing models cannot explicitly control focus planes or adjust bokeh intensity, limiting their applicability for controllable optical effects. Moreover, naively extending image-based bokeh methods to video often results in temporal flickering and unsatisfactory edge blur transitions due to the lack of temporal modeling and generalization capability. To address these challenges, we propose a novel one-step video bokeh framework that converts arbitrary input videos into temporally coherent, depth-aware bokeh effects. Our method leverages a multi-plane image (MPI) representation constructed through a progressively widening depth sampling function, providing explicit geometric guidance for depth-dependent blur synthesis. By conditioning a single-step video diffusion model on MPI layers and utilizing the strong 3D priors from pre-trained models such as Stable Video Diffusion, our approach achieves realistic and consistent bokeh effects across diverse scenes. Additionally, we introduce a progressive training strategy to enhance temporal consistency, depth robustness, and detail preservation. Extensive experiments demonstrate that our method produces high-quality, controllable bokeh effects and achieves state-of-the-art performance on multiple evaluation benchmarks. △ Less

Submitted 27 May, 2025; originally announced May 2025.

Comments: project page: https://vivocameraresearch.github.io/any2bokeh/

Concise Reasoning, Big Gains: Pruning Long Reasoning Trace with Difficulty-Aware Prompting

Authors: Yifan Wu, Jingze Shi, Bingheng Wu, Jiayi Zhang, Xiaotian Lin, Nan Tang, Yuyu Luo

Abstract: Existing chain-of-thought (CoT) distillation methods can effectively transfer reasoning abilities to base models but suffer from two major limitations: excessive verbosity of reasoning traces and inadequate adaptability to problem difficulty. Long reasoning traces significantly increase inference costs, and uniform-length solutions prevent base models from learning adaptive reasoning strategies. T… ▽ More Existing chain-of-thought (CoT) distillation methods can effectively transfer reasoning abilities to base models but suffer from two major limitations: excessive verbosity of reasoning traces and inadequate adaptability to problem difficulty. Long reasoning traces significantly increase inference costs, and uniform-length solutions prevent base models from learning adaptive reasoning strategies. To address these issues, we propose a difficulty-aware prompting (DAP) method to dynamically shorten reasoning traces without performance loss. In our approach, a large teacher model first judges each problem's difficulty and then rewrites its reasoning traces to an appropriate shorter length, yielding concise yet complete reasoning traces. Leveraging the DAP pipeline, we curate a distilled dataset called LiteCoT consisting of 100K concise reasoning examples, with solutions averaging only 720 tokens (an order of magnitude shorter than typical CoTs). Using LiteCoT, we distilled a new family of reasoning models called Liter (1.5B, 7B, and 32B) based on the Qwen2.5 architecture. Experiments show that a student model fine-tuned on just 100K of these difficulty-pruned CoT samples outperforms a model distilled on 800K original Long CoT samples, while significantly reducing training and inference costs. Our method also generalizes well: across 11 diverse benchmarks, the shorter difficulty-aware CoTs achieve equal or better accuracy than Long chains, using far fewer tokens. For example, on the challenging AIME24 exam, our approach reaches $74.2\%$ Pass@1 using only about 5K inference tokens, surpassing other methods that consume many more tokens. Our code and data are available at https://github.com/Evanwu1125/LiteCoT. △ Less

Submitted 26 May, 2025; originally announced May 2025.

Online Statistical Inference of Constrained Stochastic Optimization via Random Scaling

Authors: Xinchen Du, Wanrong Zhu, Wei Biao Wu, Sen Na

Abstract: Constrained stochastic nonlinear optimization problems have attracted significant attention for their ability to model complex real-world scenarios in physics, economics, and biology. As datasets continue to grow, online inference methods have become crucial for enabling real-time decision-making without the need to store historical data. In this work, we develop an online inference procedure for… ▽ More Constrained stochastic nonlinear optimization problems have attracted significant attention for their ability to model complex real-world scenarios in physics, economics, and biology. As datasets continue to grow, online inference methods have become crucial for enabling real-time decision-making without the need to store historical data. In this work, we develop an online inference procedure for constrained stochastic optimization by leveraging a method called Sketched Stochastic Sequential Quadratic Programming (SSQP). As a direct generalization of sketched Newton methods, SSQP approximates the objective with a quadratic model and the constraints with a linear model at each step, then applies a sketching solver to inexactly solve the resulting subproblem. Building on this design, we propose a new online inference procedure called random scaling. In particular, we construct a test statistic based on SSQP iterates whose limiting distribution is free of any unknown parameters. Compared to existing online inference procedures, our approach offers two key advantages: (i) it enables the construction of asymptotically valid confidence intervals; and (ii) it is matrix-free, i.e. the computation involves only primal-dual SSQP iterates $(\boldsymbol{x}_t, \boldsymbolλ_t)$ without requiring any matrix inversions. We validate our theory through numerical experiments on nonlinearly constrained regression problems and demonstrate the superior performance of our random scaling method over existing inference procedures. △ Less

Submitted 23 May, 2025; originally announced May 2025.

Comments: 43 pages, 1 figure, 8 tables

DiffusionReward: Enhancing Blind Face Restoration through Reward Feedback Learning

Authors: Bin Wu, Wei Wang, Yahui Liu, Zixiang Li, Yao Zhao

Abstract: Reward Feedback Learning (ReFL) has recently shown great potential in aligning model outputs with human preferences across various generative tasks. In this work, we introduce a ReFL framework, named DiffusionReward, to the Blind Face Restoration task for the first time. DiffusionReward effectively overcomes the limitations of diffusion-based methods, which often fail to generate realistic facial… ▽ More Reward Feedback Learning (ReFL) has recently shown great potential in aligning model outputs with human preferences across various generative tasks. In this work, we introduce a ReFL framework, named DiffusionReward, to the Blind Face Restoration task for the first time. DiffusionReward effectively overcomes the limitations of diffusion-based methods, which often fail to generate realistic facial details and exhibit poor identity consistency. The core of our framework is the Face Reward Model (FRM), which is trained using carefully annotated data. It provides feedback signals that play a pivotal role in steering the optimization process of the restoration network. In particular, our ReFL framework incorporates a gradient flow into the denoising process of off-the-shelf face restoration methods to guide the update of model parameters. The guiding gradient is collaboratively determined by three aspects: (i) the FRM to ensure the perceptual quality of the restored faces; (ii) a regularization term that functions as a safeguard to preserve generative diversity; and (iii) a structural consistency constraint to maintain facial fidelity. Furthermore, the FRM undergoes dynamic optimization throughout the process. It not only ensures that the restoration network stays precisely aligned with the real face manifold, but also effectively prevents reward hacking. Experiments on synthetic and wild datasets demonstrate that our method outperforms state-of-the-art methods, significantly improving identity consistency and facial details. The source codes, data, and models are available at: https://github.com/01NeuralNinja/DiffusionReward. △ Less

Submitted 23 May, 2025; originally announced May 2025.

Comments: 22 pages, 13 figures, 5 tables

NeuroTrails: Training with Dynamic Sparse Heads as the Key to Effective Ensembling

Authors: Bram Grooten, Farid Hasanov, Chenxiang Zhang, Qiao Xiao, Boqian Wu, Zahra Atashgahi, Ghada Sokar, Shiwei Liu, Lu Yin, Elena Mocanu, Mykola Pechenizkiy, Decebal Constantin Mocanu

Abstract: Model ensembles have long been a cornerstone for improving generalization and robustness in deep learning. However, their effectiveness often comes at the cost of substantial computational overhead. To address this issue, state-of-the-art methods aim to replicate ensemble-class performance without requiring multiple independently trained networks. Unfortunately, these algorithms often still demand… ▽ More Model ensembles have long been a cornerstone for improving generalization and robustness in deep learning. However, their effectiveness often comes at the cost of substantial computational overhead. To address this issue, state-of-the-art methods aim to replicate ensemble-class performance without requiring multiple independently trained networks. Unfortunately, these algorithms often still demand considerable compute at inference. In response to these limitations, we introduce $\textbf{NeuroTrails}$, a sparse multi-head architecture with dynamically evolving topology. This unexplored model-agnostic training paradigm improves ensemble performance while reducing the required resources. We analyze the underlying reason for its effectiveness and observe that the various neural trails induced by dynamic sparsity attain a $\textit{Goldilocks zone}$ of prediction diversity. NeuroTrails displays efficacy with convolutional and transformer-based architectures on computer vision and language tasks. Experiments on ResNet-50/ImageNet, LLaMA-350M/C4, among many others, demonstrate increased accuracy and stronger robustness in zero-shot generalization, while requiring significantly fewer parameters. △ Less

Submitted 23 May, 2025; originally announced May 2025.

Comments: Our open-source code is available at https://github.com/bramgrooten/neurotrails

Explain Less, Understand More: Jargon Detection via Personalized Parameter-Efficient Fine-tuning

Authors: Bohao Wu, Qingyun Wang, Yue Guo

Abstract: Personalizing jargon detection and explanation is essential for making technical documents accessible to readers with diverse disciplinary backgrounds. However, tailoring models to individual users typically requires substantial annotation efforts and computational resources due to user-specific finetuning. To address this, we present a systematic study of personalized jargon detection, focusing o… ▽ More Personalizing jargon detection and explanation is essential for making technical documents accessible to readers with diverse disciplinary backgrounds. However, tailoring models to individual users typically requires substantial annotation efforts and computational resources due to user-specific finetuning. To address this, we present a systematic study of personalized jargon detection, focusing on methods that are both efficient and scalable for real-world deployment. We explore two personalization strategies: (1) lightweight fine-tuning using Low-Rank Adaptation (LoRA) on open-source models, and (2) personalized prompting, which tailors model behavior at inference time without retaining. To reflect realistic constraints, we also investigate hybrid approaches that combine limited annotated data with unsupervised user background signals. Our personalized LoRA model outperforms GPT-4 by 21.4% in F1 score and exceeds the best performing oracle baseline by 8.3%. Remarkably, our method achieves comparable performance using only 10% of the annotated training data, demonstrating its practicality for resource-constrained settings. Our study offers the first work to systematically explore efficient, low-resource personalization of jargon detection using open-source language models, offering a practical path toward scalable, user-adaptive NLP system. △ Less

Submitted 22 May, 2025; originally announced May 2025.

Hunyuan-TurboS: Advancing Large Language Models through Mamba-Transformer Synergy and Adaptive Chain-of-Thought

Authors: Tencent Hunyuan Team, Ao Liu, Botong Zhou, Can Xu, Chayse Zhou, ChenChen Zhang, Chengcheng Xu, Chenhao Wang, Decheng Wu, Dengpeng Wu, Dian Jiao, Dong Du, Dong Wang, Feng Zhang, Fengzong Lian, Guanghui Xu, Guanwei Zhang, Hai Wang, Haipeng Luo, Han Hu, Huilin Xu, Jiajia Wu, Jianchen Zhu, Jianfeng Yan, Jiaqi Zhu , et al. (230 additional authors not shown)

Abstract: As Large Language Models (LLMs) rapidly advance, we introduce Hunyuan-TurboS, a novel large hybrid Transformer-Mamba Mixture of Experts (MoE) model. It synergistically combines Mamba's long-sequence processing efficiency with Transformer's superior contextual understanding. Hunyuan-TurboS features an adaptive long-short chain-of-thought (CoT) mechanism, dynamically switching between rapid response… ▽ More As Large Language Models (LLMs) rapidly advance, we introduce Hunyuan-TurboS, a novel large hybrid Transformer-Mamba Mixture of Experts (MoE) model. It synergistically combines Mamba's long-sequence processing efficiency with Transformer's superior contextual understanding. Hunyuan-TurboS features an adaptive long-short chain-of-thought (CoT) mechanism, dynamically switching between rapid responses for simple queries and deep "thinking" modes for complex problems, optimizing computational resources. Architecturally, this 56B activated (560B total) parameter model employs 128 layers (Mamba2, Attention, FFN) with an innovative AMF/MF block pattern. Faster Mamba2 ensures linear complexity, Grouped-Query Attention minimizes KV cache, and FFNs use an MoE structure. Pre-trained on 16T high-quality tokens, it supports a 256K context length and is the first industry-deployed large-scale Mamba model. Our comprehensive post-training strategy enhances capabilities via Supervised Fine-Tuning (3M instructions), a novel Adaptive Long-short CoT Fusion method, Multi-round Deliberation Learning for iterative improvement, and a two-stage Large-scale Reinforcement Learning process targeting STEM and general instruction-following. Evaluations show strong performance: overall top 7 rank on LMSYS Chatbot Arena with a score of 1356, outperforming leading models like Gemini-2.0-Flash-001 (1352) and o4-mini-2025-04-16 (1345). TurboS also achieves an average of 77.9% across 23 automated benchmarks. Hunyuan-TurboS balances high performance and efficiency, offering substantial capabilities at lower inference costs than many reasoning models, establishing a new paradigm for efficient large-scale pre-trained models. △ Less

Submitted 4 July, 2025; v1 submitted 21 May, 2025; originally announced May 2025.

Physics-Guided Learning of Meteorological Dynamics for Weather Downscaling and Forecasting

Authors: Yingtao Luo, Shikai Fang, Binqing Wu, Qingsong Wen, Liang Sun

Abstract: Weather forecasting is essential but remains computationally intensive and physically incomplete in traditional numerical weather prediction (NWP) methods. Deep learning (DL) models offer efficiency and accuracy but often ignore physical laws, limiting interpretability and generalization. We propose PhyDL-NWP, a physics-guided deep learning framework that integrates physical equations with latent… ▽ More Weather forecasting is essential but remains computationally intensive and physically incomplete in traditional numerical weather prediction (NWP) methods. Deep learning (DL) models offer efficiency and accuracy but often ignore physical laws, limiting interpretability and generalization. We propose PhyDL-NWP, a physics-guided deep learning framework that integrates physical equations with latent force parameterization into data-driven models. It predicts weather variables from arbitrary spatiotemporal coordinates, computes physical terms via automatic differentiation, and uses a physics-informed loss to align predictions with governing dynamics. PhyDL-NWP enables resolution-free downscaling by modeling weather as a continuous function and fine-tunes pre-trained models with minimal overhead, achieving up to 170x faster inference with only 55K parameters. Experiments show that PhyDL-NWP improves both forecasting performance and physical consistency. △ Less

Submitted 23 May, 2025; v1 submitted 20 May, 2025; originally announced May 2025.

Comments: Published/Accepted in ACM SIGKDD 2025

Empowering LLMs in Task-Oriented Dialogues: A Domain-Independent Multi-Agent Framework and Fine-Tuning Strategy

Authors: Zihao Feng, Xiaoxue Wang, Bowen Wu, Weihong Zhong, Zhen Xu, Hailong Cao, Tiejun Zhao, Ying Li, Baoxun Wang

Abstract: Task-oriented dialogue systems based on Large Language Models (LLMs) have gained increasing attention across various industries and achieved significant results. Current approaches condense complex procedural workflows into a single agent to achieve satisfactory performance on large-scale LLMs. However, these approaches face challenges to achieve comparable performance on fine-tuned lightweight LL… ▽ More Task-oriented dialogue systems based on Large Language Models (LLMs) have gained increasing attention across various industries and achieved significant results. Current approaches condense complex procedural workflows into a single agent to achieve satisfactory performance on large-scale LLMs. However, these approaches face challenges to achieve comparable performance on fine-tuned lightweight LLMs, due to their limited capabilities in handling multiple complex logic. In this work, we design a Domain-Independent Multi-Agent Framework (DIMF), which contains Intent Classification Agent, Slot Filling Agent and Response Agent. This approach simplifies the learning complexity and enhances the generalization ability by separating the tasks into domain-independent components. In this framework, we enhance the capabilities in contextual understanding using the Direct Preference Optimisation (DPO) method, and propose a simple and effective Data Distribution Adaptation (DDA) method to mitigate degradation issues during DPO training. Experiments conducted on the MultiWOZ datasets show that our proposed method achieves a better average performance among all the baselines. Extensive analysis also demonstrates that our proposed framework exhibits excellent generalizability and zero-shot capability. △ Less

Submitted 20 May, 2025; originally announced May 2025.

Dolphin: Document Image Parsing via Heterogeneous Anchor Prompting

Authors: Hao Feng, Shu Wei, Xiang Fei, Wei Shi, Yingdong Han, Lei Liao, Jinghui Lu, Binghong Wu, Qi Liu, Chunhui Lin, Jingqun Tang, Hao Liu, Can Huang

Abstract: Document image parsing is challenging due to its complexly intertwined elements such as text paragraphs, figures, formulas, and tables. Current approaches either assemble specialized expert models or directly generate page-level content autoregressively, facing integration overhead, efficiency bottlenecks, and layout structure degradation despite their decent performance. To address these limitati… ▽ More Document image parsing is challenging due to its complexly intertwined elements such as text paragraphs, figures, formulas, and tables. Current approaches either assemble specialized expert models or directly generate page-level content autoregressively, facing integration overhead, efficiency bottlenecks, and layout structure degradation despite their decent performance. To address these limitations, we present \textit{Dolphin} (\textit{\textbf{Do}cument Image \textbf{P}arsing via \textbf{H}eterogeneous Anchor Prompt\textbf{in}g}), a novel multimodal document image parsing model following an analyze-then-parse paradigm. In the first stage, Dolphin generates a sequence of layout elements in reading order. These heterogeneous elements, serving as anchors and coupled with task-specific prompts, are fed back to Dolphin for parallel content parsing in the second stage. To train Dolphin, we construct a large-scale dataset of over 30 million samples, covering multi-granularity parsing tasks. Through comprehensive evaluations on both prevalent benchmarks and self-constructed ones, Dolphin achieves state-of-the-art performance across diverse page-level and element-level settings, while ensuring superior efficiency through its lightweight architecture and parallel parsing mechanism. The code and pre-trained models are publicly available at https://github.com/ByteDance/Dolphin △ Less

Submitted 20 May, 2025; originally announced May 2025.

Comments: Accepted to ACL 2025

Smoothed SGD for quantiles: Bahadur representation and Gaussian approximation

Authors: Likai Chen, Georg Keilbar, Wei Biao Wu

Abstract: This paper considers the estimation of quantiles via a smoothed version of the stochastic gradient descent (SGD) algorithm. By smoothing the score function in the conventional SGD quantile algorithm, we achieve monotonicity in the quantile level in that the estimated quantile curves do not cross. We derive non-asymptotic tail probability bounds for the smoothed SGD quantile estimate both for the c… ▽ More This paper considers the estimation of quantiles via a smoothed version of the stochastic gradient descent (SGD) algorithm. By smoothing the score function in the conventional SGD quantile algorithm, we achieve monotonicity in the quantile level in that the estimated quantile curves do not cross. We derive non-asymptotic tail probability bounds for the smoothed SGD quantile estimate both for the case with and without Polyak-Ruppert averaging. For the latter, we also provide a uniform Bahadur representation and a resulting Gaussian approximation result. Numerical studies show good finite sample behavior for our theoretical results. △ Less

Submitted 19 May, 2025; originally announced May 2025.

BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation

Authors: Haiquan Wen, Yiwei He, Zhenglin Huang, Tianxiao Li, Zihan Yu, Xingru Huang, Lu Qi, Baoyuan Wu, Xiangtai Li, Guangliang Cheng

Abstract: Advances in AI generative models facilitate super-realistic video synthesis, amplifying misinformation risks via social media and eroding trust in digital content. Several research works have explored new deepfake detection methods on AI-generated images to alleviate these risks. However, with the fast development of video generation models, such as Sora and WanX, there is currently a lack of larg… ▽ More Advances in AI generative models facilitate super-realistic video synthesis, amplifying misinformation risks via social media and eroding trust in digital content. Several research works have explored new deepfake detection methods on AI-generated images to alleviate these risks. However, with the fast development of video generation models, such as Sora and WanX, there is currently a lack of large-scale, high-quality AI-generated video datasets for forgery detection. In addition, existing detection approaches predominantly treat the task as binary classification, lacking explainability in model decision-making and failing to provide actionable insights or guidance for the public. To address these challenges, we propose \textbf{GenBuster-200K}, a large-scale AI-generated video dataset featuring 200K high-resolution video clips, diverse latest generative techniques, and real-world scenes. We further introduce \textbf{BusterX}, a novel AI-generated video detection and explanation framework leveraging multimodal large language model (MLLM) and reinforcement learning for authenticity determination and explainable rationale. To our knowledge, GenBuster-200K is the {\it \textbf{first}} large-scale, high-quality AI-generated video dataset that incorporates the latest generative techniques for real-world scenarios. BusterX is the {\it \textbf{first}} framework to integrate MLLM with reinforcement learning for explainable AI-generated video detection. Extensive comparisons with state-of-the-art methods and ablation studies validate the effectiveness and generalizability of BusterX. The code, models, and datasets will be released. △ Less

Submitted 1 July, 2025; v1 submitted 18 May, 2025; originally announced May 2025.

LLM-CoT Enhanced Graph Neural Recommendation with Harmonized Group Policy Optimization

Authors: Hailong Luo, Bin Wu, Hongyong Jia, Qingqing Zhu, Lianlei Shan

Abstract: Graph neural networks (GNNs) have advanced recommender systems by modeling interaction relationships. However, existing graph-based recommenders rely on sparse ID features and do not fully exploit textual information, resulting in low information density within representations. Furthermore, graph contrastive learning faces challenges. Random negative sampling can introduce false negative samples,… ▽ More Graph neural networks (GNNs) have advanced recommender systems by modeling interaction relationships. However, existing graph-based recommenders rely on sparse ID features and do not fully exploit textual information, resulting in low information density within representations. Furthermore, graph contrastive learning faces challenges. Random negative sampling can introduce false negative samples, while fixed temperature coefficients cannot adapt to the heterogeneity of different nodes. In addition, current efforts to enhance recommendations with large language models (LLMs) have not fully utilized their Chain-of-Thought (CoT) reasoning capabilities to guide representation learning. To address these limitations, we introduces LGHRec (LLM-CoT Enhanced Graph Neural Recommendation with Harmonized Group Policy Optimization). This framework leverages the CoT reasoning ability of LLMs to generate semantic IDs, enriching reasoning processes and improving information density and semantic quality of representations. Moreover, we design a reinforcement learning algorithm, Harmonized Group Policy Optimization (HGPO), to optimize negative sampling strategies and temperature coefficients in contrastive learning. This approach enhances long-tail recommendation performance and ensures optimization consistency across different groups. Experimental results on three datasets demonstrate that LGHRec improves representation quality through semantic IDs generated by LLM's CoT reasoning and effectively boosts contrastive learning with HGPO. Our method outperforms several baseline models. The code is available at: https://anonymous.4open.science/r/LLM-Rec. △ Less

Submitted 18 May, 2025; originally announced May 2025.

MorphMark: Flexible Adaptive Watermarking for Large Language Models

Authors: Zongqi Wang, Tianle Gu, Baoyuan Wu, Yujiu Yang

Abstract: Watermarking by altering token sampling probabilities based on red-green list is a promising method for tracing the origin of text generated by large language models (LLMs). However, existing watermark methods often struggle with a fundamental dilemma: improving watermark effectiveness (the detectability of the watermark) often comes at the cost of reduced text quality. This trade-off limits their… ▽ More Watermarking by altering token sampling probabilities based on red-green list is a promising method for tracing the origin of text generated by large language models (LLMs). However, existing watermark methods often struggle with a fundamental dilemma: improving watermark effectiveness (the detectability of the watermark) often comes at the cost of reduced text quality. This trade-off limits their practical application. To address this challenge, we first formalize the problem within a multi-objective trade-off analysis framework. Within this framework, we identify a key factor that influences the dilemma. Unlike existing methods, where watermark strength is typically treated as a fixed hyperparameter, our theoretical insights lead to the development of MorphMarka method that adaptively adjusts the watermark strength in response to changes in the identified factor, thereby achieving an effective resolution of the dilemma. In addition, MorphMark also prioritizes flexibility since it is a model-agnostic and model-free watermark method, thereby offering a practical solution for real-world deployment, particularly in light of the rapid evolution of AI models. Extensive experiments demonstrate that MorphMark achieves a superior resolution of the effectiveness-quality dilemma, while also offering greater flexibility and time and space efficiency. △ Less

Submitted 19 May, 2025; v1 submitted 14 May, 2025; originally announced May 2025.

Comments: ACL 2025 Main

RAIDEN-R1: Improving Role-awareness of LLMs via GRPO with Verifiable Reward

Authors: Zongsheng Wang, Kaili Sun, Bowen Wu, Qun Yu, Ying Li, Baoxun Wang

Abstract: Role-playing conversational agents (RPCAs) face persistent challenges in maintaining role consistency. To address this, we propose RAIDEN-R1, a novel reinforcement learning framework that integrates Verifiable Role-Awareness Reward (VRAR). The method introduces both singular and multi-term mining strategies to generate quantifiable rewards by assessing role-specific keys. Additionally, we construc… ▽ More Role-playing conversational agents (RPCAs) face persistent challenges in maintaining role consistency. To address this, we propose RAIDEN-R1, a novel reinforcement learning framework that integrates Verifiable Role-Awareness Reward (VRAR). The method introduces both singular and multi-term mining strategies to generate quantifiable rewards by assessing role-specific keys. Additionally, we construct a high-quality, role-aware Chain-of-Thought dataset through multi-LLM collaboration, and implement experiments to enhance reasoning coherence. Experiments on the RAIDEN benchmark demonstrate RAIDEN-R1's superiority: our 14B-GRPO model achieves 88.04% and 88.65% accuracy on Script-Based Knowledge and Conversation Memory metrics, respectively, outperforming baseline models while maintaining robustness. Case analyses further reveal the model's enhanced ability to resolve conflicting contextual cues and sustain first-person narrative consistency. This work bridges the non-quantifiability gap in RPCA training and provides insights into role-aware reasoning patterns, advancing the development of RPCAs. △ Less

Submitted 15 May, 2025; originally announced May 2025.

FauForensics: Boosting Audio-Visual Deepfake Detection with Facial Action Units

Authors: Jian Wang, Baoyuan Wu, Li Liu, Qingshan Liu

Abstract: The rapid evolution of generative AI has increased the threat of realistic audio-visual deepfakes, demanding robust detection methods. Existing solutions primarily address unimodal (audio or visual) forgeries but struggle with multimodal manipulations due to inadequate handling of heterogeneous modality features and poor generalization across datasets. To this end, we propose a novel framework cal… ▽ More The rapid evolution of generative AI has increased the threat of realistic audio-visual deepfakes, demanding robust detection methods. Existing solutions primarily address unimodal (audio or visual) forgeries but struggle with multimodal manipulations due to inadequate handling of heterogeneous modality features and poor generalization across datasets. To this end, we propose a novel framework called FauForensics by introducing biologically invariant facial action units (FAUs), which is a quantitative descriptor of facial muscle activity linked to emotion physiology. It serves as forgery-resistant representations that reduce domain dependency while capturing subtle dynamics often disrupted in synthetic content. Besides, instead of comparing entire video clips as in prior works, our method computes fine-grained frame-wise audiovisual similarities via a dedicated fusion module augmented with learnable cross-modal queries. It dynamically aligns temporal-spatial lip-audio relationships while mitigating multi-modal feature heterogeneity issues. Experiments on FakeAVCeleb and LAV-DF show state-of-the-art (SOTA) performance and superior cross-dataset generalizability with up to an average of 4.83\% than existing methods. △ Less

Submitted 14 June, 2025; v1 submitted 13 May, 2025; originally announced May 2025.

Sharp Gaussian approximations for Decentralized Federated Learning

Authors: Soham Bonnerjee, Sayar Karmakar, Wei Biao Wu

Abstract: Federated Learning has gained traction in privacy-sensitive collaborative environments, with local SGD emerging as a key optimization method in decentralized settings. While its convergence properties are well-studied, asymptotic statistical guarantees beyond convergence remain limited. In this paper, we present two generalized Gaussian approximation results for local SGD and explore their implica… ▽ More Federated Learning has gained traction in privacy-sensitive collaborative environments, with local SGD emerging as a key optimization method in decentralized settings. While its convergence properties are well-studied, asymptotic statistical guarantees beyond convergence remain limited. In this paper, we present two generalized Gaussian approximation results for local SGD and explore their implications. First, we prove a Berry-Esseen theorem for the final local SGD iterates, enabling valid multiplier bootstrap procedures. Second, motivated by robustness considerations, we introduce two distinct time-uniform Gaussian approximations for the entire trajectory of local SGD. The time-uniform approximations support Gaussian bootstrap-based tests for detecting adversarial attacks. Extensive simulations are provided to support our theoretical results. △ Less

Submitted 12 May, 2025; originally announced May 2025.

BinMetric: A Comprehensive Binary Analysis Benchmark for Large Language Models

Authors: Xiuwei Shang, Guoqiang Chen, Shaoyin Cheng, Benlong Wu, Li Hu, Gangyang Li, Weiming Zhang, Nenghai Yu

Abstract: Binary analysis remains pivotal in software security, offering insights into compiled programs without source code access. As large language models (LLMs) continue to excel in diverse language understanding and generation tasks, their potential in decoding complex binary data structures becomes evident. However, the lack of standardized benchmarks in this domain limits the assessment and compariso… ▽ More Binary analysis remains pivotal in software security, offering insights into compiled programs without source code access. As large language models (LLMs) continue to excel in diverse language understanding and generation tasks, their potential in decoding complex binary data structures becomes evident. However, the lack of standardized benchmarks in this domain limits the assessment and comparison of LLM's capabilities in binary analysis and hinders the progress of research and practical applications. To bridge this gap, we introduce BinMetric, a comprehensive benchmark designed specifically to evaluate the performance of large language models on binary analysis tasks. BinMetric comprises 1,000 questions derived from 20 real-world open-source projects across 6 practical binary analysis tasks, including decompilation, code summarization, assembly instruction generation, etc., which reflect actual reverse engineering scenarios. Our empirical study on this benchmark investigates the binary analysis capabilities of various state-of-the-art LLMs, revealing their strengths and limitations in this field. The findings indicate that while LLMs show strong potential, challenges still exist, particularly in the areas of precise binary lifting and assembly synthesis. In summary, BinMetric makes a significant step forward in measuring the binary analysis capabilities of LLMs, establishing a new benchmark leaderboard, and our study provides valuable insights for the future development of these LLMs in software security. △ Less

Submitted 12 May, 2025; originally announced May 2025.

Comments: 23 pages, 5 figures, to be published in IJCAI 2025

CompileAgent: Automated Real-World Repo-Level Compilation with Tool-Integrated LLM-based Agent System

Authors: Li Hu, Guoqiang Chen, Xiuwei Shang, Shaoyin Cheng, Benlong Wu, Gangyang Li, Xu Zhu, Weiming Zhang, Nenghai Yu

Abstract: With open-source projects growing in size and complexity, manual compilation becomes tedious and error-prone, highlighting the need for automation to improve efficiency and accuracy. However, the complexity of compilation instruction search and error resolution makes automatic compilation challenging. Inspired by the success of LLM-based agents in various fields, we propose CompileAgent, the first… ▽ More With open-source projects growing in size and complexity, manual compilation becomes tedious and error-prone, highlighting the need for automation to improve efficiency and accuracy. However, the complexity of compilation instruction search and error resolution makes automatic compilation challenging. Inspired by the success of LLM-based agents in various fields, we propose CompileAgent, the first LLM-based agent framework dedicated to repo-level compilation. CompileAgent integrates five tools and a flow-based agent strategy, enabling interaction with software artifacts for compilation instruction search and error resolution. To measure the effectiveness of our method, we design a public repo-level benchmark CompileAgentBench, and we also design two baselines for comparison by combining two compilation-friendly schemes. The performance on this benchmark shows that our method significantly improves the compilation success rate, ranging from 10% to 71%. Meanwhile, we evaluate the performance of CompileAgent under different agent strategies and verify the effectiveness of the flow-based strategy. Additionally, we emphasize the scalability of CompileAgent, further expanding its application prospects. △ Less

Submitted 7 May, 2025; originally announced May 2025.

Comments: 12 pages, 4 figures

R^3-VQA: "Read the Room" by Video Social Reasoning

Authors: Lixing Niu, Jiapeng Li, Xingping Yu, Shu Wang, Ruining Feng, Bo Wu, Ping Wei, Yisen Wang, Lifeng Fan

Abstract: "Read the room" is a significant social reasoning capability in human daily life. Humans can infer others' mental states from subtle social cues. Previous social reasoning tasks and datasets lack complexity (e.g., simple scenes, basic interactions, incomplete mental state variables, single-step reasoning, etc.) and fall far short of the challenges present in real-life social interactions. In this… ▽ More "Read the room" is a significant social reasoning capability in human daily life. Humans can infer others' mental states from subtle social cues. Previous social reasoning tasks and datasets lack complexity (e.g., simple scenes, basic interactions, incomplete mental state variables, single-step reasoning, etc.) and fall far short of the challenges present in real-life social interactions. In this paper, we contribute a valuable, high-quality, and comprehensive video dataset named R^3-VQA with precise and fine-grained annotations of social events and mental states (i.e., belief, intent, desire, and emotion) as well as corresponding social causal chains in complex social scenarios. Moreover, we include human-annotated and model-generated QAs. Our task R^3-VQA includes three aspects: Social Event Understanding, Mental State Estimation, and Social Causal Reasoning. As a benchmark, we comprehensively evaluate the social reasoning capabilities and consistencies of current state-of-the-art large vision-language models (LVLMs). Comprehensive experiments show that (i) LVLMs are still far from human-level consistent social reasoning in complex social scenarios; (ii) Theory of Mind (ToM) prompting can help LVLMs perform better on social reasoning tasks. We provide some of our dataset and codes in supplementary material and will release our full dataset and codes upon acceptance. △ Less

Submitted 7 May, 2025; originally announced May 2025.

VGLD: Visually-Guided Linguistic Disambiguation for Monocular Depth Scale Recovery

Authors: Bojin Wu, Jing Chen

Abstract: Monocular depth estimation can be broadly categorized into two directions: relative depth estimation, which predicts normalized or inverse depth without absolute scale, and metric depth estimation, which aims to recover depth with real-world scale. While relative methods are flexible and data-efficient, their lack of metric scale limits their utility in downstream tasks. A promising solution is to… ▽ More Monocular depth estimation can be broadly categorized into two directions: relative depth estimation, which predicts normalized or inverse depth without absolute scale, and metric depth estimation, which aims to recover depth with real-world scale. While relative methods are flexible and data-efficient, their lack of metric scale limits their utility in downstream tasks. A promising solution is to infer absolute scale from textual descriptions. However, such language-based recovery is highly sensitive to natural language ambiguity, as the same image may be described differently across perspectives and styles. To address this, we introduce VGLD (Visually-Guided Linguistic Disambiguation), a framework that incorporates high-level visual semantics to resolve ambiguity in textual inputs. By jointly encoding both image and text, VGLD predicts a set of global linear transformation parameters that align relative depth maps with metric scale. This visually grounded disambiguation improves the stability and accuracy of scale estimation. We evaluate VGLD on representative models, including MiDaS and DepthAnything, using standard indoor (NYUv2) and outdoor (KITTI) benchmarks. Results show that VGLD significantly mitigates scale estimation bias caused by inconsistent or ambiguous language, achieving robust and accurate metric predictions. Moreover, when trained on multiple datasets, VGLD functions as a universal and lightweight alignment module, maintaining strong performance even in zero-shot settings. Code will be released upon acceptance. △ Less

Submitted 13 July, 2025; v1 submitted 5 May, 2025; originally announced May 2025.

Comments: 19 pages, conference

Adaptive Thinking via Mode Policy Optimization for Social Language Agents

Authors: Minzheng Wang, Yongbin Li, Haobo Wang, Xinghua Zhang, Nan Xu, Bingli Wu, Fei Huang, Haiyang Yu, Wenji Mao

Abstract: Effective social intelligence simulation requires language agents to dynamically adjust reasoning depth, a capability notably absent in current studies. Existing methods either lack this kind of reasoning capability or enforce Long Chain-of-Thought reasoning uniformly across all scenarios, resulting in excessive token usage and inflexible social simulation. To address this, we propose an… ▽ More Effective social intelligence simulation requires language agents to dynamically adjust reasoning depth, a capability notably absent in current studies. Existing methods either lack this kind of reasoning capability or enforce Long Chain-of-Thought reasoning uniformly across all scenarios, resulting in excessive token usage and inflexible social simulation. To address this, we propose an $\textbf{A}$daptive $\textbf{M}$ode $\textbf{L}$earning ($\textbf{AML}$) framework in this paper, aiming to improve the adaptive thinking ability of language agents in dynamic social interactions. To this end, we first identify hierarchical thinking modes ranging from intuitive response to deep deliberation based on the cognitive control theory. We then develop the $\textbf{A}$daptive $\textbf{M}$ode $\textbf{P}$olicy $\textbf{O}$ptimization ($\textbf{AMPO}$) algorithm to optimize the context-aware mode switching and reasoning. Our framework advances existing research in three key aspects: (1) Multi-granular thinking mode design, (2) Context-aware mode switching across social interaction, and (3) Token-efficient reasoning via depth-adaptive processing. Extensive experiments on social intelligence benchmarks verify that AML achieves 15.6% higher task performance than GPT-4o. Notably, our AMPO outperforms GRPO by 7.0% with 32.8% shorter reasoning chains, demonstrating the advantage of adaptive thinking mode selection and optimization mechanism in AMPO over GRPO's fixed-depth solution. △ Less

Submitted 22 May, 2025; v1 submitted 4 May, 2025; originally announced May 2025.

Comments: Work in Progress. The code and data are available, see https://github.com/MozerWang/AMPO

Robust Multi-agent Communication Based on Decentralization-Oriented Adversarial Training

Authors: Xuyan Ma, Yawen Wang, Junjie Wang, Xiaofei Xie, Boyu Wu, Shoubin Li, Fanjiang Xu, Qing Wang

Abstract: In typical multi-agent reinforcement learning (MARL) problems, communication is important for agents to share information and make the right decisions. However, due to the complexity of training multi-agent communication, existing methods often fall into the dilemma of local optimization, which leads to the concentration of communication in a limited number of channels and presents an unbalanced s… ▽ More In typical multi-agent reinforcement learning (MARL) problems, communication is important for agents to share information and make the right decisions. However, due to the complexity of training multi-agent communication, existing methods often fall into the dilemma of local optimization, which leads to the concentration of communication in a limited number of channels and presents an unbalanced structure. Such unbalanced communication policy are vulnerable to abnormal conditions, where the damage of critical communication channels can trigger the crash of the entire system. Inspired by decentralization theory in sociology, we propose DMAC, which enhances the robustness of multi-agent communication policies by retraining them into decentralized patterns. Specifically, we train an adversary DMAC\_Adv which can dynamically identify and mask the critical communication channels, and then apply the adversarial samples generated by DMAC\_Adv to the adversarial learning of the communication policy to force the policy in exploring other potential communication schemes and transition to a decentralized structure. As a training method to improve robustness, DMAC can be fused with any learnable communication policy algorithm. The experimental results in two communication policies and four multi-agent tasks demonstrate that DMAC achieves higher improvement on robustness and performance of communication policy compared with two state-of-the-art and commonly-used baselines. Also, the results demonstrate that DMAC can achieve decentralized communication structure with acceptable communication cost. △ Less

Submitted 29 April, 2025; originally announced April 2025.

Unlocking User-oriented Pages: Intention-driven Black-box Scanner for Real-world Web Applications

Authors: Weizhe Wang, Yao Zhang, Kaitai Liang, Guangquan Xu, Hongpeng Bai, Qingyang Yan, Xi Zheng, Bin Wu

Abstract: Black-box scanners have played a significant role in detecting vulnerabilities for web applications. A key focus in current black-box scanning is increasing test coverage (i.e., accessing more web pages). However, since many web applications are user-oriented, some deep pages can only be accessed through complex user interactions, which are difficult to reach by existing black-box scanners. To fil… ▽ More Black-box scanners have played a significant role in detecting vulnerabilities for web applications. A key focus in current black-box scanning is increasing test coverage (i.e., accessing more web pages). However, since many web applications are user-oriented, some deep pages can only be accessed through complex user interactions, which are difficult to reach by existing black-box scanners. To fill this gap, a key insight is that web pages contain a wealth of semantic information that can aid in understanding potential user intention. Based on this insight, we propose Hoyen, a black-box scanner that uses the Large Language Model to predict user intention and provide guidance for expanding the scanning scope. Hoyen has been rigorously evaluated on 12 popular open-source web applications and compared with 6 representative tools. The results demonstrate that Hoyen performs a comprehensive exploration of web applications, expanding the attack surface while achieving about 2x than the coverage of other scanners on average, with high request accuracy. Furthermore, Hoyen detected over 90% of its requests towards the core functionality of the application, detecting more vulnerabilities than other scanners, including unique vulnerabilities in well-known web applications. Our data/code is available at https://hoyen.tjunsl.com/ △ Less

Submitted 30 April, 2025; v1 submitted 29 April, 2025; originally announced April 2025.

WLTCL: Wide Field-of-View 3-D LiDAR Truck Compartment Automatic Localization System

Authors: Guodong Sun, Mingjing Li, Dingjie Liu, Mingxuan Liu, Bo Wu, Yang Zhang

Abstract: As an essential component of logistics automation, the automated loading system is becoming a critical technology for enhancing operational efficiency and safety. Precise automatic positioning of the truck compartment, which serves as the loading area, is the primary step in automated loading. However, existing methods have difficulty adapting to truck compartments of various sizes, do not establi… ▽ More As an essential component of logistics automation, the automated loading system is becoming a critical technology for enhancing operational efficiency and safety. Precise automatic positioning of the truck compartment, which serves as the loading area, is the primary step in automated loading. However, existing methods have difficulty adapting to truck compartments of various sizes, do not establish a unified coordinate system for LiDAR and mobile manipulators, and often exhibit reliability issues in cluttered environments. To address these limitations, our study focuses on achieving precise automatic positioning of key points in large, medium, and small fence-style truck compartments in cluttered scenarios. We propose an innovative wide field-of-view 3-D LiDAR vehicle compartment automatic localization system. For vehicles of various sizes, this system leverages the LiDAR to generate high-density point clouds within an extensive field-of-view range. By incorporating parking area constraints, our vehicle point cloud segmentation method more effectively segments vehicle point clouds within the scene. Our compartment key point positioning algorithm utilizes the geometric features of the compartments to accurately locate the corner points, providing stackable spatial regions. Extensive experiments on our collected data and public datasets demonstrate that this system offers reliable positioning accuracy and reduced computational resource consumption, leading to its application and promotion in relevant fields. △ Less

Submitted 26 April, 2025; originally announced April 2025.

Comments: To appear in IEEE TIM