1
pkg-topic-detective-noir-推理链的黑箱拆解-1786976029-1
The last time a retry loop burned through our API quota, it didn't look like a crime scene at first. It looked like a loop that had given up thinking:…
2
Calling the Bluff: Detecting Ever-Shifting Harmful Chat Dialogue via Ordered Reasoning Chain Regularization
用有序推理链正则化识别不断变化的恶意聊天内容，让有害对话无处遁形。
arXiv:2608.08451v1 Announce Type: cross Abstract: Harmful chat dialogues are ever-shifting through type-shifting and lexical evasion, yet we find they…
3
Thinking Is Not Telling: Information Disclosure in User-Service LLM Agents
大模型代理的思考过程不等于告知用户，研究揭示服务场景中推理链路的信息泄露风险。
arXiv:2602.07796v2 Announce Type: replace Abstract: User-engaged LLM agents increasingly operate in service scenarios where task success depends on co…
4
How Long Reasoning Chains Influence LLMs' Judgment of Answer Factuality
探究推理链长度如何影响大模型对答案事实性的判断，揭示关键机制。
arXiv:2604.06756v2 Announce Type: replace Abstract: Large language models (LLMs) has been widely adopted as a scalable surrogate for human evaluation,…
5
Relay-Bench: Evaluating LLMs on Multi-Domain Reasoning Chains
最新基准Relay-Bench揭示：顶级LLM在多领域推理链上仅得43.3%，尚未饱和
arXiv:2607.18438v1 Announce Type: cross Abstract: Introducing Relay-Bench, an unsaturated, holistic, text-only benchmark that measures LLMs' ability t…
6
PiCSAR: Probabilistic Confidence Selection And Ranking for Reasoning Chains
提出概率置信度选择与排序方法，优化大模型推理链，提升复杂推理的准确性与可解释性。
arXiv:2508.21787v3 Announce Type: replace-cross Abstract: Best-of-n sampling improves the accuracy of large language models (LLMs) and large reasoning…
7
Geometric Signatures of Reasoning: A Spectral Perspective on Task Hardness
从谱几何切入推理链，揭示大模型任务难度的隐藏规律，视角新颖。
arXiv:2607.01571v1 Announce Type: new Abstract: Chain-of-thought (CoT) reasoning enables large language models (LLMs) to solve complex problems by gen…
8
ReSum: Synergizing LLM Reasoning and Summarization with Reinforcement Learning
新论文ReSum用强化学习协同LLM推理与摘要，解决长推理链低效问题，干货满满。
arXiv:2606.13316v1 Announce Type: new Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) is a central technique for improving long-horizo…
9
Geometric Latent Reasoning Induces Shorter Generations in LLMs
用几何潜在推理突破传统语言链，让LLM生成更短、更便宜且不牺牲准确性，重新定义推理效率。
arXiv:2606.02248v1 Announce Type: new Abstract: Large language models solve complex problems by generating lengthy chains of explicit reasoning tokens…
10
Effective Reasoning Chains Reduce Intrinsic Dimensionality
推理链越长越聪明？新研究揭示有效推理链能显著降低任务的内在维度，ICML spotlight论文值得关注。
arXiv:2602.09276v2 Announce Type: replace-cross Abstract: Chain-of-thought (CoT) reasoning and its variants have substantially improved the performanc…
11
Can Hallucinations Be Useful? Solving Multi-Hop Questions With SLMs By Chaining System-I/II Reasoning
探讨将大型语言模型的“幻觉”转化为有用推理能力，通过链式系统I/II思维解决复杂多跳问题的新思路。
arXiv:2605.27596v1 Announce Type: new Abstract: Recently, there has been increased interest in Small Language Models (SLMs), which are fast, show good…
12
A Query Engine for the Agents
传统SQL无法查询非结构化文本，这篇论文提出专为Agent轨迹、聊天日志等数据打造的查询引擎。
arXiv:2605.27785v1 Announce Type: new Abstract: The fastest-growing data in production today is unstructured text: agent traces, chat logs, reasoning …
13
Selective Latent Thinking: Adaptive Compression of LLM Reasoning Chains
自适应压缩LLM推理链，用“选择性潜在思考”减少冗余，提升效率而保持准确性。
arXiv:2605.25745v1 Announce Type: new Abstract: Explicit chain-of-thought (CoT) reasoning substantially improves the reasoning ability of large langua…