Unverified50% confidenceFactExact time
冻结基准的设计针对AI评测领域的数据污染问题，MMLU、HumanEval等知名基准都面临数据污染困境
1
Sources
50%
Confidence
Long-term
Relevance
8/24/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedBenchmark Contamination is a significant challenge in the current field of large model evaluation, where publicly available test questions may already be included in model training data, causing evaluation results to be distorted.61% similarUnverified引导式生成模型面临数据依赖挑战，极端事件数据本就稀缺，数据质量直接决定估算可靠性61% similarUnverifiedCSAM检测AI模型面临误报率与漏报率的权衡难题：过于敏感会误拦截正常内容，过于宽松会放过违法内容60% similarUnverified业界应对大模型非确定性的测试策略包括将温度设为 0、使用 LLM-as-Judge，以及使用 Mock 数据隔离 API 调用60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/792899API
curl https://kongchang.com/api/v1/knowledge/claims/792899MCP
get_claim(id=792899)