Unverified50% confidenceFactExact time
Mainstream evaluation sets such as MMLU, HumanEval, and GSM8K have long been publicly available on the internet, making it highly likely that models have been exposed to these questions during pretraining or fine-tuning.
1
Sources
50%
Confidence
Long-term
Relevance
10/2/2026
First Seen
Sources
Related Entities
Related Claims
Unverified当前主流LLM普遍采用预训练加微调范式，先在互联网规模语料上无监督预训练，再通过RLHF等技术对齐人类偏好76% similarUnverifiedThe three major benchmarks MMLU, GSM8K, and HumanEval have all experienced varying degrees of contamination, with researchers finding content in the training data of multiple mainstream models that overlaps highly with test questions76% similarUnverifiedMainstream benchmark problems have been widely discussed on the internet, and models' training data may contain memorization of these problems, causing benchmark scores to overestimate the models' capabilities on truly unseen tasks74% similarUnverifiedArena rankings better reflect real user experience than fixed benchmarks such as GSM8K and MMLU, because fixed benchmarks are susceptible to targeted optimization and their test sets become contaminated by training data once made public.72% similarUnverifiedThe training of general-purpose large models relies on massive, publicly available, crawlable internet data.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/968799API
curl https://kongchang.com/api/v1/knowledge/claims/968799MCP
get_claim(id=968799)