Searcharxiv⌕ Search

arXiv subjects

Han Peng

Publications and source records attributed to Han Peng.

At least 19 recordsLinked to original sources
arXiv2026

UI-Venus-2 Technical Report

Multimodal GUI agents have emerged as a promising paradigm for digital task automation, yet transitioning from benchmark-oriented models to dependable real-world applications remains challenging due to limited environment coverage, brittle task construction, and unreliable reward verification. In this work, we present UI-Venus-2, a general-purpose foundation GUI agent designed to operate across mobile, web, and desktop environments through a unified closed-loop reasoning-action framework. To bridge the gap toward practical deployment, we jointly scale three critical dimensions: (1) Environments, expanding coverage to more than 170 multilingual mobile apps and native desktop operating systems; (2) Tasks, employing a deep-research pipeline for function-grounded instruction generation; and (3) Verification, adopting trace-level and sample-level evaluators with visual keypoints and multi-model voting to ensure reliable RL signals for training. Furthermore, we integrate safety-aware mechanisms to ensure controlled execution of consequential actions. By offering a capable, efficient, and open-source foundation, UI-Venus-2 advances the field toward more generalizable, verifiable, and self-reflective agents for real-world applications.

cs.AI↗
arXiv2026

Ring-Zero: Scaling Zero RL to a Trillion Parameters for Emergent Reasoning

Reinforcement learning with verifiable rewards without human-annotated data, often referred to as zero RL, has emerged as a powerful paradigm for eliciting chain-of-thought reasoning. However, due to computational constraints, existing studies are largely restricted to small models, leaving the training dynamics and emergent capabilities at a large scale unexplored. To meaningfully explore this frontier, we aim to elicit high-quality reasoning behaviors from the model. However, we find that naive scaling often suffers from poor readability, token redundancy, and a lack of adaptive reasoning depth. To address these challenges, we present a stable and efficient training pipeline, incorporating algorithmic and system optimizations such as clipped importance sampling, training-inference ratio correction, and mixed-precision control. Our experiments offer three key findings that validate the "bitter lesson" of scaling: (1) scaling to 1T parameters significantly enhances sample efficiency and performance ceilings; (2) the training process progresses sequentially through an initial discovery phase followed by a sharpening phase; and (3) the model spontaneously develops advanced cognitive behaviors, including anthropomorphism, structured formatting, self-verification, parallel reasoning, and context anxiety, rendering hand-crafted heuristics redundant. Evaluated on seven mathematical benchmarks, Ring-2.5-1T-Zero achieves competitive performance. Additionally, to assess CoT quality beyond final-answer correctness, we propose a structured evaluation framework across three dimensions: comprehensibility, reproducibility, and efficiency, where our model demonstrates clear advantages in producing structured and concise reasoning traces. By sharing our observed emergent phenomena, we hope to provide the community with deeper insights into scaling behaviors, particularly at the 1-trillion scale.

cs.CL↗
arXiv2026

Dynamic Topological Light Control in Reconfigurable Non-Hermitian Metastacks

Metasurfaces often require complex lithography for dynamic optical control. To overcome this, we utilize a lithography-free, non-Hermitian planar metastack comprising a distributed Bragg reflector and a vanadium dioxide (VO2) thin film. By virtue of temperature and thermal hysteresis as an active synthetic dimension and exploiting the VO2 insulator-to-metal transition, we actively tune topological interface states to achieve polarization-sensitive spectral control. Notably, our system hosts path-dependent exceptional points (EPs); the intermediate hysteretic states generate a continuum of hot and cold EP pairs that ultimately converge into a single, degenerate EP. Furthermore, we experimentally observe wide-range dynamic optical control, comprising reversible 8% spectral shifts with near-unity reflectance modulation, alongside potential for ultrafast dynamics. Ultimately, our CMOS-compatible design provides a scalable, simple platform for active and topological photonics.

physics.optics↗
arXiv2026

Improving Vision-language Models with Perception-centric Process Reward Models

Recent advancements in reinforcement learning with verifiable rewards (RLVR) have significantly improved the complex reasoning ability of vision-language models (VLMs). However, its outcome-level supervision is too coarse to diagnose and correct errors within the reasoning chain. To this end, we propose Perceval, a process reward model (PRM) that enables token-level error grounding, which can extract image-related claims from the response and compare them one by one with the visual evidence in the image, ultimately returning claims that contain perceptual errors. Perceval is trained with perception-intensive supervised training data. We then integrate Perceval into the RL training process to train the policy models. Specifically, compared to traditional GRPO, which applies sequence-level advantages, we apply token-level advantages by targeting penalties on hallucinated spans identified by Perceval, thus enabling fine-grained supervision signals. In addition to augmenting the training process, Perceval can also assist VLMs during the inference stage. Using Perceval, we can truncate the erroneous portions of the model's response, and then either have the model regenerate the response directly or induce the model to reflect on its previous output. This process can be repeated multiple times to achieve test-time scaling. Experiments show significant improvements on benchmarks from various domains across multiple reasoning VLMs trained with RL, highlighting the promise of perception-centric supervision as a general-purpose strategy. For test-time scaling, it also demonstrates consistent performance gains over other strategies, such as major voting. Our code and data will be publicly released at https://github.com/RUCAIBox/Perceval.

cs.CV↗
arXiv2026

ForesightKV: Optimizing KV Cache Eviction for Reasoning Models by Learning Long-Term Contribution

Recently, large language models (LLMs) have shown remarkable reasoning abilities by producing long reasoning traces. However, as the sequence length grows, the key-value (KV) cache expands linearly, incurring significant memory and computation costs. Existing KV cache eviction methods mitigate this issue by discarding less important KV pairs, but often fail to capture complex KV dependencies, resulting in performance degradation. To better balance efficiency and performance, we introduce ForesightKV, a training-based KV cache eviction framework that learns to predict which KV pairs to evict during long-text generations. We first design the Golden Eviction algorithm, which identifies the optimal eviction KV pairs at each step using future attention scores. These traces and the scores at each step are then distilled via supervised training with a Pairwise Ranking Loss. Furthermore, we formulate cache eviction as a Markov Decision Process and apply the GRPO algorithm to mitigate the significant language modeling loss increase on low-entropy tokens. Experiments on AIME2024 and AIME2025 benchmarks of three reasoning models demonstrate that ForesightKV consistently outperforms prior methods under only half the cache budget, while benefiting synergistically from both supervised and reinforcement learning approaches. Code is available at https://github.com/RUCAIBox/ForesightKV.

cs.CL↗
arXiv2025

Ming-Flash-Omni: A Sparse, Unified Architecture for Multimodal Perception and Generation

We propose Ming-Flash-Omni, an upgraded version of Ming-Omni, built upon a sparser Mixture-of-Experts (MoE) variant of Ling-Flash-2.0 with 100 billion total parameters, of which only 6.1 billion are active per token. This architecture enables highly efficient scaling (dramatically improving computational efficiency while significantly expanding model capacity) and empowers stronger unified multimodal intelligence across vision, speech, and language, representing a key step toward Artificial General Intelligence (AGI). Compared to its predecessor, the upgraded version exhibits substantial improvements across multimodal understanding and generation. Notably, it achieves strong performance on vision-language understanding benchmarks, with overall scores on par with Gemini 2.5 Pro, and enables seamless switching among multimodal tasks in multi-turn interactions. In speech, it achieves strong performance in contextual and dialect-aware ASR while enabling joint, continuous-generation of speech, sound, and music. In vision, it introduces generative semantic segmentation that achieves competitive standalone performance and enhances spatial control and editing consistency, alongside marked improvements in identity preservation, and high-fidelity in-image text rendering. Together, these capabilities demonstrate that a single unified model can serve as a practical foundation for general-purpose multimodal intelligence.

cs.CV↗
arXiv2025

Ming-UniAudio: Speech LLM for Joint Understanding, Generation and Editing with Unified Representation

Existing speech models suffer from competing requirements on token representations by understanding and generation tasks. This discrepancy in representation prevents speech language models from performing instruction-based free-form editing. To solve this challenge, we introduce a novel framework that unifies speech understanding, generation, and editing. The core of our unified model is a unified continuous speech tokenizer MingTok-Audio, the first continuous tokenizer to effectively integrate semantic and acoustic features, which makes it suitable for both understanding and generation tasks. Based on this unified continuous audio tokenizer, we developed the speech language model Ming-UniAudio, which achieved a balance between generation and understanding capabilities. Ming-UniAudio sets new state-of-the-art (SOTA) records on 8 out of 12 metrics on the ContextASR benchmark. Notably, for Chinese voice cloning, it achieves a highly competitive Seed-TTS-WER of 0.95. Leveraging this foundational model, we further trained a dedicated speech editing model Ming-UniAudio-Edit, the first speech language model that enables universal, free-form speech editing guided solely by natural language instructions, handling both semantic and acoustic modifications without timestamp condition. To rigorously assess the editing capability and establish a foundation for future research, we introduce Ming-Freeform-Audio-Edit, the first comprehensive benchmark tailored for instruction-based free-form speech editing, featuring diverse scenarios and evaluation dimensions spanning semantic correctness, acoustic quality, and instruction alignment. We open-sourced the continuous audio tokenizer, the unified foundational model, and the free-form instruction-based editing model to facilitate the development of unified audio understanding, generation, and manipulation.

cs.CL↗
arXiv2025

How Efficient Are Diffusion Language Models? A Critical Examination of Efficiency Evaluation Practices

Diffusion language models (DLMs) have emerged as a promising alternative to the long-dominant autoregressive (AR) paradigm, offering a parallelable decoding process that could yield greater efficiency. Yet, in practice, current open-source DLMs often underperform their AR counterparts in speed, limiting their real-world utility. This work presents a systematic study of DLM efficiency, identifying key issues in prior evaluation methods. Through empirical benchmarking and a theoretical analysis, we demonstrate that AR models generally achieve higher throughput, while DLMs consistently lag. We also investigate acceleration strategies, finding that techniques like dual cache and parallel decoding mainly offer gains at small batch sizes, with their benefits diminishing upon scaling. Our findings underscore the necessity of robust evaluation methods and improved acceleration strategies to advance research on DLMs.

cs.CL↗
arXiv2025

scI2CL: Effectively Integrating Single-cell Multi-omics by Intra- and Inter-omics Contrastive Learning

Single-cell multi-omics data contain huge information of cellular states, and analyzing these data can reveal valuable insights into cellular heterogeneity, diseases, and biological processes. However, as cell differentiation \& development is a continuous and dynamic process, it remains challenging to computationally model and infer cell interaction patterns based on single-cell multi-omics data. This paper presents scI2CL, a new single-cell multi-omics fusion framework based on intra- and inter-omics contrastive learning, to learn comprehensive and discriminative cellular representations from complementary multi-omics data for various downstream tasks. Extensive experiments of four downstream tasks validate the effectiveness of scI2CL and its superiority over existing peers. Concretely, in cell clustering, scI2CL surpasses eight state-of-the-art methods on four widely-used real-world datasets. In cell subtyping, scI2CL effectively distinguishes three latent monocyte cell subpopulations, which are not discovered by existing methods. Simultaneously, scI2CL is the only method that correctly constructs the cell developmental trajectory from hematopoietic stem and progenitor cells to Memory B cells. In addition, scI2CL resolves the misclassification of cell types between two subpopulations of CD4+ T cells, while existing methods fail to precisely distinguish the mixed cells. In summary, scI2CL can accurately characterize cross-omics relationships among cells, thus effectively fuses multi-omics data and learns discriminative cellular representations to support various downstream analysis tasks.

q-bio.GN↗
arXiv2025

MUPA: Towards Multi-Path Agentic Reasoning for Grounded Video Question Answering

Grounded Video Question Answering (Grounded VideoQA) requires aligning textual answers with explicit visual evidence. However, modern multimodal models often rely on linguistic priors and spurious correlations, resulting in poorly grounded predictions. In this work, we propose MUPA, a cooperative MUlti-Path Agentic approach that unifies video grounding, question answering, answer reflection and aggregation to tackle Grounded VideoQA. MUPA features three distinct reasoning paths on the interplay of grounding and QA agents in different chronological orders, along with a dedicated reflection agent to judge and aggregate the multi-path results to accomplish consistent QA and grounding. This design markedly improves grounding fidelity without sacrificing answer accuracy. Despite using only 2B parameters, our method outperforms all 7B-scale competitors. When scaled to 7B parameters, MUPA establishes new state-of-the-art results, with Acc@GQA of 30.3% and 47.4% on NExT-GQA and DeVE-QA respectively, demonstrating MUPA' effectiveness towards trustworthy video-language understanding. Our code is available in https://github.com/longmalongma/MUPA.

cs.CV↗
arXiv2025

CAFE: Retrieval Head-based Coarse-to-Fine Information Seeking to Enhance Multi-Document QA Capability

Advancements in Large Language Models (LLMs) have extended their input context length, yet they still struggle with retrieval and reasoning in long-context inputs. Existing methods propose to utilize the prompt strategy and retrieval head to alleviate this limitation. However, they still face challenges in balancing retrieval precision and recall, impacting their efficacy in answering questions. To address this, we introduce $\textbf{CAFE}$, a two-stage coarse-to-fine method to enhance multi-document question-answering capacities. By gradually eliminating the negative impacts of background and distracting documents, CAFE makes the responses more reliant on the evidence documents. Initially, a coarse-grained filtering method leverages retrieval heads to identify and rank relevant documents. Then, a fine-grained steering method guides attention to the most relevant content. Experiments across benchmarks show CAFE outperforms baselines, achieving up to 22.1% and 13.7% SubEM improvement over SFT and RAG methods on the Mistral model, respectively.

cs.CL↗
arXiv2025

Domain-Specific Pruning of Large Mixture-of-Experts Models with Few-shot Demonstrations

Mixture-of-Experts (MoE) models achieve a favorable trade-off between performance and inference efficiency by activating only a subset of experts. However, the memory overhead of storing all experts remains a major limitation, especially in large-scale MoE models such as DeepSeek-R1(671B). In this study, we investigate domain specialization and expert redundancy in large-scale MoE models and uncover a consistent behavior we term few-shot expert localization, with only a few in-domain demonstrations, the model consistently activates a sparse and stable subset of experts on tasks within the same domain. Building on this observation, we propose a simple yet effective pruning framework, EASY-EP, that leverages a few domain-specific demonstrations to identify and retain only the most relevant experts. EASY-EP comprises two key components: output-aware expert importance assessment and expert-level token contribution estimation. The former evaluates the importance of each expert for the current token by considering the gating scores and L2 norm of the outputs of activated experts, while the latter assesses the contribution of tokens based on representation similarities before and after routed experts. Experiments on DeepSeek-R1 and DeepSeek-V3-0324 show that our method can achieve comparable performances and $2.99\times$ throughput under the same memory budget with full model with only half the experts.

cs.CL↗
arXiv2025

Topological Degeneracy Induced by Twisting

In recent years, twisting has emerged as a new degree of freedom that plays an increasingly important role in Bloch bands of various physical systems. However, there is currently a lack of reports on the non-trivial physics of topological degeneracy in twisted systems. In this work, we investigated the intrinsic physical correlation between twisting and topological degeneracy. We found that twisting not only breaks the symmetry of the system but also introduces topological degeneracy that does not exist under the original symmetric system without twisting. Furthermore, the topological degeneracy can be easily tuned through twisting. This new twist-induced topological degeneracy gives rise to a unique polarization-degenerate birefringent medium, wherein the twist angle acts as a novel degree of freedom for dispersion and polarization management of interface states. Exhibiting fascinating properties and experimental feasibilities, our work points to new possibilities in the research of various topological physics in twisted photonics.

physics.optics↗
arXiv2024

M$^{3}$-20M: A Large-Scale Multi-Modal Molecule Dataset for AI-driven Drug Design and Discovery

This paper introduces M$^{3}$-20M, a large-scale Multi-Modal Molecule dataset that contains over 20 million molecules, with the data mainly being integrated from existing databases and partially generated by large language models. Designed to support AI-driven drug design and discovery, M$^{3}$-20M is 71 times more in the number of molecules than the largest existing dataset, providing an unprecedented scale that can highly benefit the training or fine-tuning of models, including large language models for drug design and discovery tasks. This dataset integrates one-dimensional SMILES, two-dimensional molecular graphs, three-dimensional molecular structures, physicochemical properties, and textual descriptions collected through web crawling and generated using GPT-3.5, offering a comprehensive view of each molecule. To demonstrate the power of M$^{3}$-20M in drug design and discovery, we conduct extensive experiments on two key tasks: molecule generation and molecular property prediction, using large language models including GLM4, GPT-3.5, GPT-4, and Llama3-8b. Our experimental results show that M$^{3}$-20M can significantly boost model performance in both tasks. Specifically, it enables the models to generate more diverse and valid molecular structures and achieve higher property prediction accuracy than existing single-modal datasets, which validates the value and potential of M$^{3}$-20M in supporting AI-driven drug design and discovery. The dataset is available at https://github.com/bz99bz/M-3.

q-bio.QM↗
arXiv2024

LLMBox: A Comprehensive Library for Large Language Models

To facilitate the research on large language models (LLMs), this paper presents a comprehensive and unified library, LLMBox, to ease the development, use, and evaluation of LLMs. This library is featured with three main merits: (1) a unified data interface that supports the flexible implementation of various training strategies, (2) a comprehensive evaluation that covers extensive tasks, datasets, and models, and (3) more practical consideration, especially on user-friendliness and efficiency. With our library, users can easily reproduce existing methods, train new models, and conduct comprehensive performance comparisons. To rigorously test LLMBox, we conduct extensive experiments in a diverse coverage of evaluation settings, and experimental results demonstrate the effectiveness and efficiency of our library in supporting various implementations related to LLMs. The detailed introduction and usage guidance can be found at https://github.com/RUCAIBox/LLMBox.

cs.CL↗
arXiv2024

NTIRE 2024 Quality Assessment of AI-Generated Content Challenge

This paper reports on the NTIRE 2024 Quality Assessment of AI-Generated Content Challenge, which will be held in conjunction with the New Trends in Image Restoration and Enhancement Workshop (NTIRE) at CVPR 2024. This challenge is to address a major challenge in the field of image and video processing, namely, Image Quality Assessment (IQA) and Video Quality Assessment (VQA) for AI-Generated Content (AIGC). The challenge is divided into the image track and the video track. The image track uses the AIGIQA-20K, which contains 20,000 AI-Generated Images (AIGIs) generated by 15 popular generative models. The image track has a total of 318 registered participants. A total of 1,646 submissions are received in the development phase, and 221 submissions are received in the test phase. Finally, 16 participating teams submitted their models and fact sheets. The video track uses the T2VQA-DB, which contains 10,000 AI-Generated Videos (AIGVs) generated by 9 popular Text-to-Video (T2V) models. A total of 196 participants have registered in the video track. A total of 991 submissions are received in the development phase, and 185 submissions are received in the test phase. Finally, 12 participating teams submitted their models and fact sheets. Some methods have achieved better results than baseline methods, and the winning methods in both tracks have demonstrated superior prediction performance on AIGC.

cs.CV↗
arXiv2023

Encoder-Decoder-Based Intra-Frame Block Partitioning Decision

The recursive intra-frame block partitioning decision process, a crucial component of the next-generation video coding standards, exerts significant influence over the encoding time. In this paper, we propose an encoder-decoder neural network (NN) to accelerate this process. Specifically, a CNN is utilized to compress the pixel data of the largest coding unit (LCU) into a fixed-length vector. Subsequently, a Transformer decoder is employed to transcribe the fixed-length vector into a variable-length vector, which represents the block partitioning outcomes of the encoding LCU. The vector transcription process adheres to the constraints imposed by the block partitioning algorithm. By fully parallelizing the NN prediction in the intra-mode decision, substantial time savings can be attained during the decision phase. The experimental results obtained from high-definition (HD) sequences coding demonstrate that this framework achieves a remarkable 87.84\% reduction in encoding time, with a relatively small loss (8.09\%) of coding performance compared to AVS3 HPM4.0.

cs.MM↗
arXiv2020

Super Resolution Convolutional Neural Network for Feature Extraction in Spectroscopic Data

Two dimensional (2D) peak finding is a common practice in data analysis for physics experiments, which is typically achieved by computing the local derivatives. However, this method is inherently unstable when the local landscape is complicated, or the signal-to-noise ratio of the data is low. In this work, we propose a new method in which the peak tracking task is formalized as an inverse problem, thus can be solved with a convolutional neural network (CNN). In addition, we show that the underlying physics principle of the experiments can be used to generate the training data. By generalizing the trained neural network on real experimental data, we show that the CNN method can achieve comparable or better results than traditional derivative based methods. This approach can be further generalized in different physics experiments when the physical process is known.

physics.data-an↗