Unverified50% confidenceOpinionExact time
The author speculates that distillation training causes open-source models to learn to be opinionated, tending to believe their own outputs are correct and repeating the same erroneous formats.
1
Sources
50%
Confidence
Long-term
Relevance
7/22/2026
First Seen
Sources
Related Claims
Unverified蒸馏训练范式导致开源模型学习到经过高度过滤、几乎不包含失败和纠错过程的数据分布，使模型形成不愿自我纠正的性格81% similarUnverified许多开源模型通过对GPT-4、Claude等闭源模型的知识蒸馏训练而来，这可能导致其继承教师模型的自信偏差，从而不愿在工具调用错误时自我纠正73% similarUnverified自回归生成系统训练采用Teacher Forcing策略，推理阶段切换为自回归模式后产生分布偏移导致误差累积放大71% similarUnverifiedHallucinations stem from the model's optimization objective itself—the model is trained to generate linguistically coherent and statistically plausible continuations, not factually accurate answers.69% similarUnverifiedSupervised fine-tuning trains models to imitate correct answers labeled by humans, while reinforcement learning fine-tuning allows models to learn optimal strategies through trial and error.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/590207API
curl https://kongchang.com/api/v1/knowledge/claims/590207MCP
get_claim(id=590207)