Unverified50% confidenceConceptExact time
Automatic conversation evaluation (LLM-as-a-Judge) is the mainstream approach to agent evaluation today: using another large model to score conversation records according to preset scoring criteria, or to judge whether replies satisfy dimensions such as factuality, tone, and compliance.
1
Sources
50%
Confidence
Long-term
Relevance
11/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLLM-as-a-judge evaluates Agent behavior by combining the Agent's execution trace and evaluation criteria into a prompt that is submitted to an independent judge model for scoring or pass/reject decisions.76% similarUnverified"LLM as judge" is one of the mainstream paradigms for evaluating generative AI systems today, and its basic approach is to have a more capable language model score or rank the outputs of the system under test according to preset grading criteria.74% similarVerifiedLLM as Judge is a technique that uses an independent judge LLM to evaluate the output quality of the main model, with evaluation dimensions including safety, relevance, and factual consistency.71% similarVerifiedLLM-as-a-judge模式即用一个能力更强的模型来评判另一个模型的输出质量，降低评测成本但引入评判模型自身的偏见问题70% similarUnverifiedSome teams have introduced an LLM-as-Judge approach, using another model to evaluate agent output quality in order to form an automated quality feedback loop70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1123655API
curl https://kongchang.com/api/v1/knowledge/claims/1123655MCP
get_claim(id=1123655)