Unverified50% confidenceTradeoffExact time
Muon-RT's convergence guarantee relies on an alignment assumption, namely that the inner product between the teacher displacement and the current gradient is non-negative; if the teacher lags too long and deviates severely from the student's current direction, the assumption may be violated.
1
Sources
50%
Confidence
Long-term
Relevance
11/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWeak-to-strong generalization does not hold universally, but depends on conditions such as target structure and training methodology; when conditions are not met, the student may fail to surpass the teacher70% similarUnverifiedSimple token removal can only weaken the transfer effect of the teacher's safety tendencies to the student and cannot completely eliminate it; post-hoc DPO alignment can also only partially correct this imprinted behavioral tendency.69% similarUnverifiedThe paper finds that when the student's trajectory moves in the wrong direction, the teacher does not correct the error but instead continues writing along the student's error, that is, the teacher has the ability to recognize errors but lacks the ability to escape from them.69% similarUnverifiedThe paper "Trajectory as the Teacher" proposes that the true bottleneck is the trajectory itself rather than the student model.67% similarVerifiedThe teacher model outputs in distillation training rarely include self-correction examples, causing student models to form an implicit belief that "strong models don't make mistakes," leading them to repeat rather than correct when verification fails.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1125154API
curl https://kongchang.com/api/v1/knowledge/claims/1125154MCP
get_claim(id=1125154)