Unverified50% confidenceOpinionExact time
The problem with AGENTS.md has the same origin as the system prompt: the model may not read it, may forget it after reading it, and after forgetting may still act according to its default habits from training, so its effectiveness is probabilistic.
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhen endowing existing VLMs with generative capabilities, the understanding abilities accumulated during model pretraining are often impaired, resulting in catastrophic forgetting.70% similarUnverified训练样本末尾遗漏EOS token会导致模型无法学习何时停止生成，推理时陷入无限续写70% similarVerifiedBehavioral cloning simplifies the problem to supervised learning, but suffers from distribution shift, where differences between the state distributions at training time and execution time lead to error accumulation.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1018274API
curl https://kongchang.com/api/v1/knowledge/claims/1018274MCP
get_claim(id=1018274)