Unverified50% confidenceFactExact time
Benchmarks commonly used by the open-source community (such as MMLU, HumanEval, MATH, and LongBench) carry risks of data contamination.
1
Sources
50%
Confidence
Long-term
Relevance
9/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIf a malicious dataset were uploaded to a public platform and adopted by downstream users, its potential harm would be amplified along the model training and deployment chain, posing a supply chain security risk to the open-source ecosystem.67% similarUnverifiedFor scenarios where databases are exposed to public network queries via Datasette, the priority for security upgrades should be higher.65% similarUnverifiedUsing real-time web crawling as a core differentiator introduces potential risks including data compliance, crawling stability, and copyright attribution of referenced content.65% similarUnverifiedMeta、谷歌等开源路线支持者认为开源能通过社区审查提升安全性并防止少数实验室垄断AI能力65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/962918API
curl https://kongchang.com/api/v1/knowledge/claims/962918MCP
get_claim(id=962918)