Unverified50% confidenceOpinionExact time
Unlike directly applying cross-entropy (hard labels) to the correction segments, KL supervision uses the teacher's full soft distribution as a soft target, preserving the teacher's relative confidence across candidate tokens, yielding smoother signals and making it less likely for the student to overfit to a single answer.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedToken-level supervision means that the teacher gives a distribution or label for each token position in the sequence, and the student learns by minimizing the difference from the teacher's distribution (such as KL divergence).73% similarUnverifiedTeacher intervention pilot studies show that teacher preferences differ across tokens, and that through majority voting and KL-divergence-based selection, a better teacher can be identified for each token.71% similarVerifiedModel distillation uses soft labels from a large teacher model to train a small student model; the dark knowledge carried in the soft labels makes distillation more efficient than ordinary supervised learning.70% similarUnverifiedSeTPO introduces teacher signals only for low-reward trajectories, uses reverse KL divergence to derive token-level teacher feedback, and incorporates it into the policy advantage function.69% similarUnverifiedOn-policy distillation requires the student to first generate its own prefix trajectory, after which the teacher provides guidance at the states the student actually reaches, so the state distributions are aligned, and it is usually more stable on long-horizon tasks where errors tend to accumulate.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1115283API
curl https://kongchang.com/api/v1/knowledge/claims/1115283MCP
get_claim(id=1115283)