Verified65% confidenceFactExact time
When a GPT-4-level model acts as a Judge, its evaluation results agree with human experts more than 80% of the time.
3
Sources
65%
Confidence
Long-term
Relevance
3/2/2026
First Seen
Sources
Cursor 2.0 in Practice: Vibe Coding Methodology & Guide to 10x Coding Efficiency
bilibili认真的笨笨3/2/2026
Related Entities
Related Claims
Unverified研究表明GPT-4作为评估者与人类专家评分的一致性超过80%，显著高于BLEU、ROUGE等基于词汇匹配的自动化指标84% similarUnverifiedGPT-4等模型在作为评判者时会系统性地偏好特定位置的回答77% similarUnverifiedRepresentative works of the "LLM-as-a-judge" paradigm include MT-Bench/Chatbot Arena by Zheng et al., who found that GPT-4 as a judge achieves over 80% agreement with human preferences.75% similarUnverifiedOpenAI与Nature合作评估了GPT-4的审稿能力，发现其与人类审稿人的意见重合度约为50-60%74% similarVerifiedThe LLM-as-Judge method uses large language models such as GPT-4 as judges to automatically score the output quality of Agents.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/2105API
curl https://kongchang.com/api/v1/knowledge/claims/2105MCP
get_claim(id=2105)