Wenxuan Huang
Author directoryOther people with similar names: Wenxuan Huang
2026
Stable-MM-R1: Anchoring Multimodal Reasoning Dynamics via Entropy-Guided Stratification
Yimeng Ye | Shuang Chen | Wenxuan Huang | Manyuan Zhang | Kaituo Feng | Zhangquan Chen | Jiayu Chen | Yucheng Zhou | YiCheng Xiao | ZhiYuan Feng | Tianyu Shi
Findings of the Association for Computational Linguistics: EMNLP 2026
Yimeng Ye | Shuang Chen | Wenxuan Huang | Manyuan Zhang | Kaituo Feng | Zhangquan Chen | Jiayu Chen | Yucheng Zhou | YiCheng Xiao | ZhiYuan Feng | Tianyu Shi
Findings of the Association for Computational Linguistics: EMNLP 2026
While Reinforcement Learning (RL) effectively incentivizes reasoning in Large Language Models, current pipelines are hindered by training instability and rapid entropy collapse. These limitations often stem from "Rollout Silencing" and low-quality gradient signals in standard sampling procedures. In this work, we propose a robust, data-centric framework to stabilize RL training. We first introduce Potential-Aware Query Mining (PAQM), which filters data dynamically to focus on the "Distillation Zone"—samples with high potential for capability elicitation. Furthermore, we present Hybrid Stratified Replay (HSR), a novel mechanism that restructures batches by stratifying rollouts based on Path Entropy, a rollout-level confidence proxy, and outcome reward. Within each optimization step, HSR reuses current-policy "Stability Anchors" and "Hard Negatives" to construct high-contrast optimization groups, then clears its buffers before the next step. This approach mitigates entropy collapse while improving the utilization of learning signals under limited compute. Our method outperforms strong baselines on complex reasoning tasks, offering a principled solution for stable and efficient RL fine-tuning.
Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent
Zhen Fang | Yu Zeng | Wenxuan Huang | Yiming Zhao | Shiting Huang | Tianfei Ren | Qi Lu | Qingnan Ren | Qisheng Su | Lionel Z. Wang | Qingyu Yin | Shuang Chen | Zehui Chen | Lin Chen | Zhenfei Yin | Yao Hu | Shaohui Lin | Wanli Ouyang | Shaosheng Cao | Feng Zhao
Findings of the Association for Computational Linguistics: EMNLP 2026
Zhen Fang | Yu Zeng | Wenxuan Huang | Yiming Zhao | Shiting Huang | Tianfei Ren | Qi Lu | Qingnan Ren | Qisheng Su | Lionel Z. Wang | Qingyu Yin | Shuang Chen | Zehui Chen | Lin Chen | Zhenfei Yin | Yao Hu | Shaohui Lin | Wanli Ouyang | Shaosheng Cao | Feng Zhao
Findings of the Association for Computational Linguistics: EMNLP 2026
We introduce Video-DeepResearch (Video-DR), extending multimodal agents from static images to continuous video streams, a setting that demands dense spatiotemporal grounding coupled with open-web exploration. Preliminary evaluations reveal two critical bottlenecks in current models: (1) modality bias, where agents bypass visual tools in favor of textual search, and (2) parametric knowledge leakage, where models rely on internal memory rather than genuine tool-augmented execution. To address these challenges, we propose Video-DeepResearch, featuring a decoupled perception-exploration pipeline with stage-wise tool unlocking that compels exhaustive cross-frame visual grounding prior to web retrieval. Our framework adopts a two-stage training recipe—supervised fine-tuning followed by Group Relative Policy Optimization (GRPO)—enabling autonomous exploration that breaks the imitation-learning ceiling. Furthermore, we curate VideoDR-Bench, a human-AI collaborative benchmark comprising 200 complex, multi-hop VQA instances. Empirical results demonstrate that our Video-DeepResearch-35B-A3B establishes a new state-of-the-art of 64.0% average accuracy, surpassing proprietary Claude-4.5-Sonnet (59.0%) by 5.0 points and significantly outperforming GPT-5 (52.5%) and Gemini 2.5 Pro (57.5%). The 30B-A3B variant achieves 59.3%, competitive with Claude-4.5-Sonnet and demonstrating the effectiveness of our training paradigm even at compact scale.
LLaVA-RadZ: Can Multimodal Large Language Models Effectively Tackle Zero-shot Radiology Recognition?
Bangyan Li | Wenxuan Huang | Zhenkun Gao | Yeqiang Wang | Yuxiang Shen
Findings of the Association for Computational Linguistics: EMNLP 2026
Bangyan Li | Wenxuan Huang | Zhenkun Gao | Yeqiang Wang | Yuxiang Shen
Findings of the Association for Computational Linguistics: EMNLP 2026
Recently, Multimodal Large Language Models (MLLMs) have demonstrated strong visual understanding and reasoning abilities across vision-language tasks. However, conventional Visual Question Answering (VQA) pipelines do not fully exploit the visual features and medical knowledge captured by MLLMs, leading to limited performance in zero-shot medical disease recognition. To address this issue, we propose LLaVA-RadZ, a simple yet effective framework for zero-shot radiology recognition. LLaVA-RadZ introduces Decoding-Side Feature Alignment Training (DFAT), which leverages decoder-side representations with modality-specific tokens for image-text alignment, and a Domain Knowledge Anchoring Module (DKAM), which incorporates disease-aware semantic anchors to mitigate the category semantic gap. Extensive experiments show that LLaVA-RadZ substantially improves the zero-shot disease recognition ability of MLLMs and achieves competitive performance compared with well-established CLIP-based radiology recognition methods.
Rethinking Diffusion Large Language Models via Time-aware Weighting and Block-oriented Refinements
Zheyong Xie | Shengrui Li | Li Zhong | Kaiyan Zhao | Fangcheng Shi | Jiakang Wang | Huayi Jin | Zhaokai Luo | Fei Zhao | Haifeng Liu | Jieying Ye | Chonggang Lu | Wenxuan Huang | Yue Liu | Zijie Meng | Shaosheng Cao
Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
Zheyong Xie | Shengrui Li | Li Zhong | Kaiyan Zhao | Fangcheng Shi | Jiakang Wang | Huayi Jin | Zhaokai Luo | Fei Zhao | Haifeng Liu | Jieying Ye | Chonggang Lu | Wenxuan Huang | Yue Liu | Zijie Meng | Shaosheng Cao
Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
Diffusion Large Language Models (DLLMs) have recently emerged as a parallel alternative to autoregressive models, offering bidirectional context modeling and parallel decoding. However, developing highly performant DLLMs is hindered by lifecycle challenges, including suboptimal timestep weighting in pre-training, train-inference distribution gaps in post-training, and underutilized caching during inference. To address these obstacles, we introduce ReDLLM, a 30B-parameter Mixture-of-Experts DLLM. It incorporates a time-aware weighting strategy during continual pre-training that scales gradients according to denoising difficulty. Furthermore, we propose two block-oriented refinements: a terminal-block treatment to narrow the train-inference gap, and a block-aware inference pipeline featuring inline block prefill and block-level scheduling to maximize KV-cache reuse and parallelization. Extensive evaluations demonstrate that ReDLLM surpasses existing DLLMs at the same active-parameter scale, achieves generation quality competitive with 100B DLLMs, and attains the highest inference throughput among evaluated models.
UniCorn: Towards Self-Improving Unified Multimodal Models through Self-Generated Supervision
Zhen Fang | Ruiyan Han | XinYu Sun | Yuchen Ma | Ziheng Wang | Yu Zeng | Zehui Chen | Lin Chen | Wenxuan Huang | Wei-Jie Xu | Yi Cao | Feng Zhao
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Zhen Fang | Ruiyan Han | XinYu Sun | Yuchen Ma | Ziheng Wang | Yu Zeng | Zehui Chen | Lin Chen | Wenxuan Huang | Wei-Jie Xu | Yi Cao | Feng Zhao
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
While Unified Multimodal Models (UMMs) have achieved remarkable success in cross-modal comprehension, a significant gap persists in their ability to leverage such internal knowledge for high-quality generation. We formalize this discrepancy as Conduction Aphasia, a phenomenon where models accurately interpret multimodal inputs but struggle to translate that understanding into faithful and controllable synthesis. To address this, we propose UniCorn, a simple yet elegant self-improvement framework that eliminates the need for external data or teacher supervision. By partitioning a single UMM into three collaborative roles: Proposer, Solver, and Judge, UniCorn generates high-quality interactions via self-play and employs cognitive pattern reconstruction to distill latent understanding into explicit generative signals. To validate the restoration of multimodal coherence, we introduce UniCycle, a cycle-consistency benchmark based on a Text to Image to Text reconstruction loop. Extensive experiments demonstrate that UniCorn achieves comprehensive and substantial improvements over the base model across six general image generation benchmarks. Notably, it achieves SOTA performance on TIIF(73.8), DPG(86.8), CompBench(88.5), and UniCycle while further delivering substantial gains of +5.0 on WISE and +6.5 on OneIG. These results highlight that our method significantly enhances T2I generation while maintaining robust comprehension, demonstrating the scalability of fully self-supervised refinement for unified multimodal intelligence.
MT3: A Synergistic Multi-Task RL Framework for Specializing MLLMs in Text Image Machine Translation
Zhaopeng Feng | Yupu Liang | Shaosheng Cao | Jiayuan Su | Jiahan Ren | Zhijie Zhou | Wenxuan Huang | Jian Wu | Zuozhu Liu
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Zhaopeng Feng | Yupu Liang | Shaosheng Cao | Jiayuan Su | Jiahan Ren | Zhijie Zhou | Wenxuan Huang | Jian Wu | Zuozhu Liu
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Text Image Machine Translation (TIMT)—the task of translating textual content embedded in images—is critical for applications in accessibility, cross-lingual information access, and real-world document understanding. However, TIMT remains a complex challenge due to the need for accurate optical character recognition (OCR), robust visual-text reasoning, and high-quality translation, often requiring cascading multi-stage pipelines. Recent advances in large-scale Reinforcement Learning (RL) have improved reasoning in Large Language Models (LLMs) and Multimodal LLMs (MLLMs), but their application to end-to-end TIMT is still underexplored. To bridge this gap, we introduce MT3, a novel Multi-Task RL framework to specialize MLLMs into end-to-end expert TIMT models. MT3 adopts a synergistic multi-task optimization paradigm targeting three key sub-skills: text recognition, context-aware reasoning, and translation. It is trained using a novel multi-mixed reward mechanism that provides fine-grained feedback, fostering a controllable and transparent optimization process. Furthermore, to facilitate the evaluation of TIMT in authentic cross-cultural and real-world social media contexts, we introduced XHSPost, the first social media TIMT benchmark. Our MT3-7B-Zero achieves state-of-the-art results on the latest in-domain MIT-10M benchmark, outperforming strong baselines such as Qwen2.5-VL-72B and InternVL2.5-78B by notable margins across multiple metrics. Additionally, the model shows strong generalization to out-of-distribution language pairs and datasets. In-depth analyses reveal how multi-task synergy, reinforcement learning initialization, curriculum design, and reward formulation contribute to advancing MLLM-driven TIMT.
Search
Co-authors
- Shaosheng Cao 3
- Lin Chen 2
- Shuang Chen 2
- Zehui Chen 2
- Zhen Fang 2
- Yu Zeng 2
- Feng Zhao 2
- Yi Cao 1
- Jiayu Chen 1
- Zhangquan Chen 1
- Kaituo Feng 1
- Zhaopeng Feng 1
- ZhiYuan Feng 1
- Zhenkun Gao 1
- Ruiyan Han 1
- Yao Hu 1
- Shiting Huang 1
- Huayi Jin 1
- Bangyan Li 1
- Shengrui Li 1
- Yupu Liang 1
- Shaohui Lin 1
- Haifeng Liu 1
- Yue Liu 1
- Zuozhu Liu 1
- Chonggang Lu 1
- Qi Lu 1
- Zhaokai Luo 1
- Yuchen Ma 1
- Zijie Meng 1
- Wanli Ouyang 1
- Jiahan Ren 1
- Qingnan Ren 1
- Tianfei Ren 1
- Yuxiang Shen 1
- Fangcheng Shi 1
- Tianyu Shi 1
- Jiayuan Su 1
- Qisheng Su 1
- XinYu Sun 1
- Jiakang Wang 1
- Lionel Z. Wang 1
- Yeqiang Wang 1
- Ziheng Wang 1
- Jian Wu 1
- Yicheng Xiao 1
- Zheyong Xie 1
- Wei-Jie Xu 1
- Jieying Ye 1
- Yimeng Ye 1
- Qingyu Yin 1
- Zhenfei Yin 1
- Manyuan Zhang 1
- Fei Zhao 1
- Kaiyan Zhao 1
- Yiming Zhao 1
- Li Zhong 1
- Yucheng Zhou 1
- Zhijie Zhou 1