Unverified50% confidenceFactExact time
The study proposes that the true reason velocity scaling is effective is population time lag.
1
Sources
50%
Confidence
Long-term
Relevance
10/9/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe core motivation behind the deceleration argument is that the rate of model capability growth is outpacing humans' ability to understand and constrain those capabilities.76% similarUnverifiedVelocity scaling and shifting the model time backward are essentially two equivalent approaches to addressing the population time lag problem.75% similarUnverifiedProgress in alignment research is widely considered to lag far behind the pace of improvement in model capabilities themselves.70% similarUnverifiedPopulation time lag refers to the phenomenon where states sampled at model time t more closely resemble the state distribution at an earlier time point in the training data.67% similarUnverifiedThe previously dominant view held that velocity fields trained with mean squared error (MSE) systematically underestimate velocity magnitudes, and that velocity scaling is a correction for this error.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/990724API
curl https://kongchang.com/api/v1/knowledge/claims/990724MCP
get_claim(id=990724)