Unverified50% confidenceDecision RuleExact time
The judge itself is also a component that needs version management—replacing the judge model or modifying the rubric makes historical scores incomparable, so baseline experiments must be recorded bound to the judge version.
1
Sources
50%
Confidence
Long-term
Relevance
10/29/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe version-coupling problem of LLM-as-judge: the judge model itself is also updated with versions, and its scoring criteria and tolerance for wording may shift; when the evaluated model and the judge model are upgraded at the same time, changes in evaluation results are hard to attribute.74% similarUnverifiedIf the judge is changed frequently, historical scores lose comparability and the trend line breaks.73% similarUnverifiedIn non-stationary environments, the candidate superiority relations reflected by previously collected labels may become invalid, and a controller that clings to old evidence will make outdated decisions.72% similarUnverifiedEvaluation scores rely on test datasets, and as datasets change with business needs, score drift occurs, making historical cross-sectional comparisons misleading.70% similarUnverifiedOne practical approach to handling delayed ground truth: directly rewrite the evaluation record of the original run once the ground truth arrives; the advantage is a single source of truth, and the disadvantage is that it breaks the immutability of the evaluation history.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1113039API
curl https://kongchang.com/api/v1/knowledge/claims/1113039MCP
get_claim(id=1113039)