揭示大模型优势源于约束引导推理，而非单纯规模扩大，颠覆传统认知。
arXiv:2606.26108v1 Announce Type: new Abstract: Larger language models consistently outperform smaller ones on reasoning benchmarks, yet the reasoning…
探讨人类与LLM日常推理中共享的模式匹配机制，挑战“真正推理”与“模式匹配”的二分法
arXiv:2606.13607v1 Announce Type: new Abstract: When large language models (LLMs) fail to generalize or make haphazard errors in reasoning, it is ofte…
在《我的世界》中实现长期规划，一项名为WISE的智能体通过“原因-选择”推理突破复杂任务，为AI自主决策提供新范式。
arXiv:2606.12852v1 Announce Type: new Abstract: Rapid advances have been made in developing general-purpose embodied agent in environments like Minecr…
ICML 2026论文揭示LLM推理时注意力引导的信息流，为强化学习提供精准干预目标。
arXiv:2606.10646v1 Announce Type: new Abstract: Token-level credit assignment remains a key obstacle for reinforcement learning (RL) in large language…
揭示大语言模型推理时被表面启发式误导而忽略隐含约束的机制，对理解AI的认知偏差有重要启发。
arXiv:2603.29025v3 Announce Type: replace-cross Abstract: Large language models fail when a salient surface cue conflicts with an unstated feasibility…
Claude与ChatGPT的"思考投入"机制是独立模型还是同参数不同效果？作者通过缓存行为差异切入，探讨大模型推理能力的实现方式。
Claude and ChatGPT have thinking efforts where you can tune the amount of thinking allowed. Like low, medium, high, xhigh and so on. But are they diff…
ICML2026接收论文，用熵动力学揭示Chain-of-Thought推理的深层机理，为理解大模型思维链过程提供新视角。
arXiv:2606.02020v1 Announce Type: cross Abstract: This paper investigates the entropy dynamics of Chain-of-Thought (CoT) and uncovers a consistent two…
ICML 2026论文揭示：LLM代码理解靠先验知识，而非编程语言语义
arXiv:2510.03415v3 Announce Type: replace-cross Abstract: Recent work asks whether large language models (LLMs) condition their reasoning on explicit …
从线性化表示角度拆解大模型在结构化知识上产生幻觉的深层机理，ACL 2026论文揭示推理缺陷。
arXiv:2605.26362v1 Announce Type: cross Abstract: In many reasoning tasks, large language models (LLMs) rely on structured external knowledge, such as…
探秘LLM推理的黑箱：通过机械性解释揭示不同语义结构如何中介推理过程。
arXiv:2605.25520v1 Announce Type: new Abstract: Predicting a label correctly does not necessarily require representing the operation that produces it.…
ICML 2026 接收论文，将 k-means 聚类重新诠释为链式思维推理，为图学习提供全新数学视角。
arXiv:2605.24867v1 Announce Type: new Abstract: Chain-of-Thought (CoT) prompting has shown promise in enhancing the reasoning capabilities of large la…
最新arXiv论文揭秘思维链如何一步步拆解复杂任务，深入剖析其内在分解机制。
arXiv:2604.08872v2 Announce Type: replace Abstract: Many language tasks can be modeled as classification problems where a large language model (LLM) i…
用符号图建模多智能体间的冲突信号，提升复杂协作场景下的推理鲁棒性，开辟AI冲突管理新路径。
arXiv:2605.19418v1 Announce Type: new Abstract: LLM-based multi-agent systems (MAS) have demonstrated strong reasoning and decision-making capabilitie…