9 papers
MechReason: Benchmarking Multi-Image Multi-Hop Reasoning in Mechanical Engineering
Tengyue Wang, Kang An, Chenxu Du +8
Despite significant progress in general visual question answering and cross-modal understanding, multimodal large language models still face a pronounced gap in evaluation for comp…
CircuitReason-1k: Benchmarking Long-Horizon Visual-to-Symbolic Reasoning inElectrical Circuits
Xinqi Yang, Kang An, Tengyue Wang +8
Electrical circuit analysis requires more than recognizing components in an image. A solver must ground symbols and labels, recover latent topology, select a physical model, formul…
MMArch: Benchmarking Multimodal Reasoning Grounded in Architectural Evidence
Chenxu Du, Kang An, Tengyue Wang +8
Multimodal large language models (MLLMs) perform strongly on engineering imagery, yet existing benchmarks mostly test drawing recognition, information extraction, or compliance che…
SafeSceneReason: A Multimodal Reasoning Benchmark Connecting Industrial Hazards with Accident Knowledge
Yuanchi Zhu, Kang An, Tengyue Wang +11
Industrial-safety understanding requires more than detecting workers, equipment, and personal protective equipment. Models must also assess compliance, identify hazardous interacti…
Shattering the Autoregressive Curse: Dynamic Epistemic Entropy Orchestrated Erasable Reinforcement Learning for LLMs
Ziliang Wang, Kang An, Faqiang Qian +5
Although reinforcement learning (RL) has expanded the cognitive boundaries of large language models (LLMs), it often remains vulnerable to the autoregressive curse in long-horizon…
SELF-EMO: Emotional Self-Evolution from Recognition to Consistent Expression
Shaowei Zhang, Faqiang Qian, Yan Chen +5
Emotion Recognition in Conversation (ERC) has become a fundamental capability for large language models (LLMs) in human-centric interaction. Beyond accurate recognition, coherent e…
MMRPT: MultiModal Reinforcement Pre-Training via Masked Vision-Dependent Reasoning
Xuhui Zheng, Kang An, Ziliang Wang +3
Multimodal pre-training remains constrained by the descriptive bias of image-caption pairs, leading models to favor surface linguistic cues over grounded visual understanding. We i…
Erase to Improve: Erasable Reinforcement Learning for Search-Augmented LLMs
Ziliang Wang, Kang An, Xuhui Zheng +6
While search-augmented large language models (LLMs) exhibit impressive capabilities, their reliability in complex multi-hop reasoning remains limited. This limitation arises from t…
AAPA: Adversarially Anchored Preference Alignment for Post-Training of Large Language Models
Faqiang Qian, Kang An, Weikun Zhang +6
Post-training alignment of large language models often combines supervised fine-tuning (SFT) on expert demonstrations with reinforcement learning (RL) from preference or verifiable…