Unverified50% confidenceFactExact time
When supervision is weak, lightweight SFT can reinforce the teacher's context-based refusal behavior, significantly improving both the SS1 and SS0.5 scores.
1
Sources
50%
Confidence
Long-term
Relevance
11/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThrough targeted teacher intervention, SSD avoids the low-signal state that readily arises in full-policy rollouts, because the teacher provides signals only at positions where knowledge injection is genuinely needed.75% similarUnverifiedCompared with the approach of directly applying supervised fine-tuning (SFT) followed by reinforcement learning, the rubric-guided method performs better.74% similarUnverifiedSFT is usually much cheaper than pretraining but is extremely sensitive to data quality; data ratios, the number of epochs, and the learning rate all significantly affect results.72% similarUnverifiedExperimental results show that SAPD improves performance under different teacher configurations and effectively mitigates catastrophic forgetting compared with SFT.72% similarUnverifiedSupervised fine-tuning (SFT) optimizes imitation loss rather than a behavioral safety objective, so it readily exhibits the phenomenon of "working within the training distribution but failing out of distribution."71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1179603API
curl https://kongchang.com/api/v1/knowledge/claims/1179603MCP
get_claim(id=1179603)