Theory of Mind (ToM) refers to the ability to infer others' mental states, which is an essential capability for embodied AI agents to effectively collaborate and interact with humans. While improving Large Language Models' ability to reason about characters' mental states in text-based stories/dialogues has been extensively studied, enhancing Multimodal Large Language Models' ToM capabilities, particularly in egocentric video from an embodied perspective, remains unexplored. In this paper, we propose a contrastive Reinforcement Learning (RL) paradigm that explicitly encourages models to leverage temporal and causal evolutionary patterns in user action sequences to infer user's mental states (goals, beliefs, and potential next actions). Evaluation results on in-domain and out-of-domain demonstrate that our method achieves performance improvements of (+30.00%, +2.00%) and (+5.83%, +5.00%) compared to the backbone model and vanilla Group Relative Policy Optimization (GRPO) model, respectively. Additionally, we compare the performance of two post-training paradigms (Supervise Fine-Tuning and RL) and systematically analyze the reasoning trajectories across the base model, vanilla GRPO model, and our proposed method.

Teng Pan, Yuchen Yan, Zixuan Wang, Ruiqing Zhang, Guiyang Hou, Wenqi Zhang, Weiming Lu, Jun Xiao, Yongliang Shen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Wenqi Zhang, Mengna Wang, Gangao Liu, Huixin Xu, Yiwei Jiang, Yongliang Shen, Guiyang Hou, Zhe Zheng, Hang Zhang, Xin Li, Jiajun Liu, Weiming Lu, Peng Li, Yueting Zhuang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Social intelligence rests on three foundational pillars: cognitive intelligence, situational intelligence, and behavioral intelligence. As Large Language Model (LLM)-based Agents are increasingly integrated into our social life, understanding, evaluating, and advancing their social intelligence has emerged as a problem of considerable importance. Although a growing body of work has investigated the social intelligence of LLMs, several critical gaps remain: (1) Existing studies predominantly concentrate on a single pillar, while a comprehensive framework for organizing and studying the social intelligence of LLMs is still lacking; (2) existing evaluations position LLMs as passive observers from a third-person perspective. Compared to the third-person perspective, egocentric first-person evaluation more faithfully reflects the deployment conditions of LLM-based agents in practice; (3) there is a lack of human-LLM interaction paradigms for evaluating behavioral intelligence, which would enable more intuitive comparison of behavioral differences between humans and LLM Agents. In light of these limitations, we introduce Social 3 -EGO, a comprehensive egocentric evaluation framework grounded in three foundational pillars of social intelligence, incorporating four key technical contributions: strategies for perspective shifting, synthesis of diverse real-world situational scenarios, construction of adversarial agents via rules and RL training, as well as new rubrics for human-LLM interaction evaluation. Using Social 3 -EGO, we conduct an evaluation of fourteen frontier foundation models, revealing a substantial gap from human-level social intelligence.

Large Language Models (LLMs) and Multimodal LLMs have shown promising capabilities for SVG processing, yet existing benchmarks suffer from limited real-world coverage, lack of complexity stratification, and fragmented evaluation paradigms. We introduce SVGenius, a comprehensive benchmark comprising 2,377 queries across three progressive dimensions: understanding, editing, and generation. Built on real-world data from 24 application domains with systematic complexity stratification, SVGenius evaluates models through 8 task categories and 18 metrics. We assess 22 mainstream models spanning different scales, architectures, training paradigms, and accessibility levels. Our analysis reveals that while proprietary models significantly outperform open-source counterparts, all models exhibit systematic performance degradation with increasing complexity, indicating fundamental limitations in current approaches; however, reasoning-enhanced training proves more effective than pure scaling for overcoming these limitations, though style transfer remains the most challenging capability across all model types. SVGenius establishes the first systematic evaluation framework for SVG processing, providing crucial insights for developing more capable vector graphics models and advancing automated graphic design applications. Appendix and supplementary materials (including all data and code) are available at https://zju-real.github.io/SVGenius.

Large language models (LLMs) have achieved remarkable progress on mathematical tasks through Chain-of-Thought (CoT) reasoning. However, existing mathematical CoT datasets often suffer from **Thought Leaps** due to experts omitting intermediate steps, which negatively impacts model learning and generalization. We propose the CoT Thought Leap Bridge Task, which aims to automatically detect leaps and generate missing intermediate reasoning steps to restore the completeness and coherence of CoT. To facilitate this, we constructed a specialized training dataset called **ScaleQM+**, based on the structured ScaleQuestMath dataset, and trained **CoT-Bridge** to bridge thought leaps. Through comprehensive experiments on mathematical reasoning benchmarks, we demonstrate that models fine-tuned on bridged datasets consistently outperform those trained on original datasets, with improvements of up to +5.87\% on NuminaMath. Our approach effectively enhances distilled data (+3.02\%) and provides better starting points for reinforcement learning (+3.1\%), functioning as a plug-and-play module compatible with existing optimization techniques. Furthermore, CoT-Bridge demonstrates improved generalization to out-of-domain logical reasoning tasks, confirming that enhancing reasoning completeness yields broadly applicable benefits.

Wenqi Zhang, Ke Tang, Hai Wu, Mengna Wang, Yongliang Shen, Guiyang Hou, Zeqi Tan, Peng Li, Yueting Zhuang, Weiming Lu. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Wenqi Zhang, Zhenglin Cheng, Yuanyu He, Mengna Wang, Yongliang Shen, Zeqi Tan, Guiyang Hou, Mingqian He, Yanna Ma, Weiming Lu, Yueting Zhuang. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Math Solving requires both semantic understanding and relation reasoning. Most current approaches treat it as a translation task from natural language to mathematical symbols, generating tokens one by one. However, token-level generation is usually vulnerable when confronted with diverse annotations and complex reasoning. We consider the equation is an ordered combination of multiple sub-expressions, and math reasoning should be performed on the sub-expression level rather than the token level. We treat sub-expression as a minimum generative unit and minimum reasoning node. At each step, candidate sub-expression nodes are generated in parallel, and the whole reasoning chain is deduced by combining multiple nodes in order. Besides, we can obtain multiple valid reasoning chains by sub-expression searching, further improving interpretability and precision. Experiments on multilingual datasets show our method significantly outperforms the baselines. Additionally, our approach is more stable and efficient when faced with the challenges of diverse annotation and complex reasoning with limited resources. Moreover, our approach can be seamlessly integrated with large language models (LLMs), enhancing LLM's mathematical reasoning capabilities at a minimal cost. Experiments show a synergistic collaboration between general-purpose LLM and our specialized model yields superior performance.