1
智谱提前剧透GLM-6.0：完全自训练方法公开了
模型、论文都还没出，没想到先从财务公告上剧透了
2
When AI Reviews Its Own Code: Recursive Self-Training Collapse in Code LLMs
代码大模型陷入自训练怪圈：这篇论文揭示了AI自我审查代码时可能发生的递归塌缩现象。
arXiv:2606.28438v1 Announce Type: cross Abstract: Recursive self-training can degrade neural generative models when generated data is reused without f…
3
PASTA: A Paraphrasing And Self-Training Approach for Knowledge Updating in LLMs
用改写加自训练，搞定大模型的知识更新难题，方法新颖实用。
arXiv:2606.28898v1 Announce Type: new Abstract: Knowledge updating in pre-trained Large Language Models (LLMs) remains an important challenge. While c…
4
Be My Tutor: On-Policy Co-Distillation for Mutual LLM Improvement via Peer Feedback
LLM互教互学新范式：通过同伴反馈实现策略共蒸馏，让两个模型在训练中相互提升。
arXiv:2606.14368v1 Announce Type: new Abstract: We study multi-domain LLM training in which two models, each stronger in a different domain, co-evolve…
5
Bi-CoG: Bi-Consistency-Guided Self-Training for Vision-Language Models
提出双一致性引导自训练策略，从特征与预测两个维度对齐，显著增强视觉语言模型的零样本泛化能力。
arXiv:2510.20477v3 Announce Type: replace Abstract: Exploiting unlabeled data through semi-supervised learning (SSL) or leveraging pre-trained models …
6
Found in Conversation: LLMs Teach Themselves to Close the Multi-Turn Gap
研究发现LLM在多轮对话中表现远逊于单轮，但模型能通过自训练弥补这一差距，值得AI研究者关注。
arXiv:2605.24432v1 Announce Type: new Abstract: Large Language Model (LLM) interactions are typically underspecified, with users clarifying all necess…