Unverified50% confidenceFactExact time
The test concluded that stuffing the context with documents is not the real risk; what truly drags down smaller models is cross-document reasoning.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/12/2026
First Seen
Valid until: 1/10/2027
Sources
Related Entities
Related Claims
Unverified漏洞数量激增并不必然意味着模型更不安全，可能是审查强度提升带来的统计偏差72% similarVerifiedSome models carry the risk of over-optimization toward test sets, causing leaderboard scores to deviate from real-world performance (benchmark saturation).71% similarUnverifiedOnce a public benchmark attracts widespread attention, it risks being targeted for optimization, and high scores do not necessarily reflect generalized reasoning ability.71% similarUnverifiedVersion drift is dangerous in evaluation scenarios because the validity of historical measurement results implicitly decays after changes to model weights, inference frameworks, or test environments70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1003217API
curl https://kongchang.com/api/v1/knowledge/claims/1003217MCP
get_claim(id=1003217)