Unverified50% confidenceFactExact time
Outcome-aware likelihood comparison evaluates skill utility by computing the log-likelihood of the model reproducing the trajectory under the two conditions of carrying and not carrying the candidate skill, and weighting and comparing the likelihood difference according to the trajectory outcome label.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedDuring model selection, it is necessary to confirm that the similarity objective function used during model training is consistent with the metric used during retrieval; otherwise, systematic bias will occur.77% similarUnverifiedVariance-based contrastive optimization identifies problems in planning and tool invocation by contrasting samples with different performance levels.76% similarUnverifiedA more rigorous approach to evaluating genuine model progress is to conduct controlled comparative tests while progressively reducing dependence on scaffolding.74% similarUnverifiedEvaluating coding agent tools should probe whether their performance is stable across different underlying models; a significant performance drop after swapping out the model indicates the harness is highly dependent on a specific model.72% similarUnverifiedGraftOOD adopts a shared velocity-regression objective function, combining two types of supervision signals: group-relative reward correction and sequence-level example supervision.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1028033API
curl https://kongchang.com/api/v1/knowledge/claims/1028033MCP
get_claim(id=1028033)