Unverified50% confidenceFactExact time
研究表明，超过80%的审稿人在rebuttal后不会实质性更改评分
1
Sources
50%
Confidence
Long-term
Relevance
7/11/2026
First Seen
Sources
The Pain of AI Paper Peer Review: Why Should Models Released After the Deadline Count Against You?
redditr/LanguageTechnology7/9/2026
Related Claims
Unverified多项针对ACL、EMNLP、ICLR等会议的统计分析显示，rebuttal导致审稿人实质性提分的概率通常低于20%67% similarUnverifiedResearch shows that the agreement rate between LLM judges and human scores is approximately 80%, approaching but not yet surpassing the level of human inter-annotator agreement (approximately 81%).64% similarVerifiedThe LLM-as-Judge method can achieve over 80% agreement with human expert scoring.63% similarUnverifiedResearch shows that when humans assign 90% confidence, their actual accuracy rate is often only around 70%, indicating a systematic overconfidence61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/476190API
curl https://kongchang.com/api/v1/knowledge/claims/476190MCP
get_claim(id=476190)