Unverified50% confidenceFactExact time
In the standard SFT process, the model continuously expends gradients on sequences that are already well-fitted, causing the output distribution to narrow and entropy to decrease, thereby compressing the exploration space for subsequent RL.
1
Sources
50%
Confidence
Long-term
Relevance
9/7/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedExcessively thorough SFT may harm the exploration space for subsequent RL, and a perfectly converged SFT model is not necessarily the best starting point for RL.71% similarUnverifiedSFT fine-tuning is effective but causes a decline in general knowledge in small models—i.e., the phenomenon of catastrophic forgetting—due to the prominent capacity bottleneck of small models.69% similarUnverified基于过程变量的闭环控制属于条件维护策略，相比定时维护能显著降低资源消耗68% similarUnverified通过'不要按常规套路'这样的约束性提示词，本质上是在调整模型的采样分布，降低高频方案的输出权重，迫使模型探索低概率但可能更具创新性的输出区域68% similarUnverifiedCross-entropy loss often converges faster than MSE in classification tasks because, when paired with softmax output, gradients do not vanish in the saturation region.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/873212API
curl https://kongchang.com/api/v1/knowledge/claims/873212MCP
get_claim(id=873212)