Unverified50% confidenceOpinionExact time
The metric truly worth tracking is cost per correct answer, because it is the only metric that cannot be optimized by failing more quietly.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe cost-per-correct-answer metric simultaneously penalizes both extremes: high failure rate but cheap, and high success rate but expensive.74% similarUnverifiedWhen multiple models are tied on accuracy, the cost per flawless run becomes the truly useful tiebreaker metric.74% similarUnverifiedThe number of valid issues caught per dollar should be used as the core evaluation metric, rather than accuracy alone.71% similarUnverified仅用小模型策略的每正确答案成本最低($0.0004)但准确率仅33%，说明没有准确率底线的每正确答案成本指标具有误导性71% similarUnverifiedCost per correct answer is economically equivalent to unit output cost, a standard efficiency metric in manufacturing and advertising (e.g., CPM, CPA), but has not yet become convention in AI evaluation frameworks.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1003282API
curl https://kongchang.com/api/v1/knowledge/claims/1003282MCP
get_claim(id=1003282)