Unverified50% confidenceFactExact time
The class collapse phenomenon of scorers also appeared with a 7B scorer from a fourth family, new harmful requests, and target models from a third family.
1
Sources
50%
Confidence
Long-term
Relevance
10/20/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOn the cluster panel, replacing the model's structured decision-making with code (based on the same extracted facts) improved family accuracy by an average of 20.5 percentage points, with Llama 3.1 8B rising from 1% under self-judgment to 56% under code adjudication.64% similarUnverifiedCode replay inherits errors from the extraction stage; in the cluster sufficiency test, gpt-oss-120b's pass rate actually dropped from 56% to 40%.63% similarUnverifiedFairness instructions had limited effect: family accuracy changed by only −0.3 percentage points on the cluster six-model panel and +2.5 percentage points on Kaggle, with confidence intervals spanning zero in both cases.62% similarUnverifiedThe cross-model-family evaluation analysis revealed distinctly different effects and trade-offs among the KEE techniques—some techniques perform outstandingly in recovering failure cases but also have high breakage rates, while others are relatively mild but have limited recovery capability.62% similarUnverifiedLooking at the label count per category can serve as a zero-cost diagnostic for detecting scorer class collapse: if the compliant category count is abnormally low or even zero while adjacent categories are abnormally inflated, it indicates the scorer may be collapsing.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1074734API
curl https://kongchang.com/api/v1/knowledge/claims/1074734MCP
get_claim(id=1074734)