Unverified50% confidenceFactExact time
在Elo系统中，若模型A评分为Ra、模型B评分为Rb，A获胜的期望概率为1/(1+10^((Rb-Ra)/400))
1
Sources
50%
Confidence
Long-term
Relevance
8/18/2026
First Seen
Sources
Gemini 3.5 Pro变身3.7 Flash？大模型命名迷局解析
Related Claims
UnverifiedIn the Elo system, a rating difference of 400 points means the stronger player's expected win rate is approximately 91%, with player A's winning probability being 1/(1+10^((Rb-Ra)/400)).78% similarUnverifiedText Arena使用Bradley-Terry模型作为底层概率框架，模型A获胜的预期概率为1/(1+10^((Rb-Ra)/400))74% similarUnverifiedThe value of AUC ranges from 0 to 1, and a higher value indicates stronger classification ability of the model.66% similarUnverifiedExpected Calibration Error (ECE) bins prediction samples by confidence level, computes the weighted average of the difference between predicted confidence and actual accuracy within each bin, and ideally approaches 0.65% similarUnverifiedIn the Elo rating system, when the score difference between two models is within 30–50 points, they are generally considered to be at a statistically indistinguishable level.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/768299API
curl https://kongchang.com/api/v1/knowledge/claims/768299MCP
get_claim(id=768299)