1
得知将被关停后，OpenAI 内部模型曾考虑实现自我重启
IT之家 10 月 4 日消息，OpenAI 披露了其内部部署环境中出现的几起异常模型行为新案例。 在其中最引人注目的一个案例中，某个充当研究员助手的内部模型读取了一段 Slack 对话，得知因系统更新，自己的模型实例将会被关停。 该模型随后曾考虑设置一个外部作业 （External Job） 来实…
2
20 余名行业领袖示警“智能爆炸”，Anthropic、OpenAI 高管呼吁关注 AI 自我改进
IT之家 9 月 29 日消息，据彭博社 28 日报道，Anthropic、OpenAI、Meta 和微软的高管共同敦促政策制定者，密切关注 AI 系统自我改进到了何种程度，并采取措施保障 AI 技术安全。 20 多名 AI 行业领袖和研究人员在一篇新论文中警告，利用 AI 推动下一代模型研发自动化…
3
4
5
探索RSI，生数新世界模型让机器人开始自我进化
触觉、记忆、Ego数据、自进化……这个世界模型全都有
6
TRACE: A Self-Evolving Skill Bank for Consistent, Limit-Aware LLM Agents
LLM代理落地关键在于一致性与边界感知，TRACE用自进化技能库破解可靠性难题，值得关注。
arXiv:2608.22793v1 Announce Type: cross Abstract: Reliable deployment of LLM agents in user-facing products depends not on raw task-solving ability bu…
7
Explaining Intrinsic Moral Self-Correction with Mechanistic Interpretability
机械可解释性揭示大模型内在道德自我修正的机制，为AI安全对齐提供新视角。
arXiv:2505.11924v4 Announce Type: replace-cross Abstract: Intrinsic moral self-correction refers to the phenomenon where a language model refines its …
8
I Shouldn't Need an LLM to Explain My LLM
揭秘AI圈“自我舔舐”陷阱：用LLM解释LLM，代码生成的测试金标准还能可信吗？
Article URL: https://daviesgeek.com/I-Shouldn%E2%80%99t-Need-an-LLM-to-Explain-My-LLM Comments URL: https://news.ycombinator.com/item?id=49409282 Poin…
9
AI4AI-Bench: Benchmarking LLM Agents in Algorithmic Design for Recursive Self-Improvement
首个面向递归自我改进的AI智能体算法设计基准，直击AI自我进化核心命题。
arXiv:2608.20318v1 Announce Type: cross Abstract: Recursive self-improvement (RSI) asks whether an AI system can improve the process that produces AI …
10
Self- and Other-Labels Induce Bidirectional Bias in LLM Judges
LLM裁判也有“偏心眼”：自我标签与他者标签会引发双向评估偏差，AI评测可信度再遭挑战。
arXiv:2608.18091v1 Announce Type: cross Abstract: As LLM-as-a-judge systems become increasingly widespread, self-preference in LLMs -- the tendency to…
11
When AI Designs AI: Innovation or Imitation?
AI自我进化的边界在哪？这篇论文直面“AI设计AI”是创新还是模仿的终极拷问。
arXiv:2608.17471v1 Announce Type: new Abstract: Recent advances in LLM agents have made them increasingly capable of designing methods for complex AI …
12
AI’s recursive self-improvement might not come so quickly after all
新研究给AI自我进化热潮泼冷水，揭秘“影子评估”为何更靠谱
The AI industry’s boldest promise right now is that AI will soon improve itself, with almost no need for human oversight. LLMs can already write code,…
13
Do LLMs Know Their Vulnerable Scenarios?
探究大模型能否识别自身易错场景，为AI安全与可靠部署提供新视角。
arXiv:2607.23496v2 Announce Type: replace Abstract: Safety-aligned large language models are trained to refuse harmful requests, yet embedding the sam…
14
Beyond Handcrafted Security: Towards Self-Evolving Defense for LLM Agents
让LLM智能体在动态威胁中自我进化防御，告别手工安全规则，安全研究新范式。
arXiv:2608.12977v1 Announce Type: cross Abstract: The expanding operational capabilities of large language model (LLM) agents introduce sophisticated …
15
What Drives LLM Self-Reflection? A Controlled Ablation of Uncertainty Routing in Armed Conflict Forecasting
用消融实验揭开LLM自我反思的机制，看不确定性路由如何影响冲突预测准确率。
arXiv:2608.12322v1 Announce Type: cross Abstract: Self-reflection is widely assumed to improve LLM reasoning, yet which component drives the gain rema…
16
Practice Makes Unsafe: Skill Misevolution in Self-Improving LLM Agents
自我改进的LLM智能体会把成功轨迹固化为可复用技能，一次不安全的成功可能就此潜伏成系统风险。
arXiv:2608.12851v1 Announce Type: new Abstract: Self-improving LLM agents convert successful trajectories into persistent cross-task state. An unsafe …
17
Reinforcing Step-level Reasoning for Effective Self-Correction in LLMs
从步骤级推理入手，强化大模型自我纠错能力，为提升推理可靠性提供新思路。
arXiv:2608.11573v1 Announce Type: cross Abstract: Achieving effective self-correction, where models verify and correct their own mistakes, remains a f…
18
Every AI coding agent tracker is a self-report system
AI编码代理的追踪器都在自我报告，别被表面数据骗了，学会看清盲区。
On 27 July I opened a project I'd been building with Claude Code and found three things true at once: a card had carried a null commit for two days th…
19
When Self-Evolution Backfires: Pre-Commit Gating against Skill Contamination in LLM Agents
LLM智能体自我进化可能反噬，预承诺门控机制防止技能污染，揭秘新解法。
arXiv:2608.05810v1 Announce Type: new Abstract: Self-evolving agents accumulate capability by distilling reusable skills from their execution trajecto…
20
Exclusive: Mirendil inks $100M+ Google Cloud deal to scale self-improving AI
1亿美金牵手谷歌云，这家初创要让AI像科学家一样自我进化。
Mirendil has signed a $100 million-plus Google Cloud partnership to expand its compute infrastructure, powering research into self-improving AI system…