Unverified50% confidenceFactExact time
Only differential contamination—where some models are contaminated far more than others—distorts rankings, and this situation is rare in practice.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified研究表明即使是部分污染（模型只见过题目而非答案），也可能显著提升得分74% similarUnverified研究表明即使是轻微的数据污染也可能导致评测分数虚高5-15个百分点69% similarUnverifiedOut of a total of 188 model × benchmark combinations, only 3 cases exhibit differential contamination cross-validated by two reference points.69% similarUnverifiedRecent research indicates that some models perform 10–15 percentage points higher on known contaminated benchmark subsets than on uncontaminated subsets.69% similarPartially VerifiedBenchmark contamination refers to a model having already seen test questions during training, resulting in inflated scores that fail to reflect true generalization ability.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/900514API
curl https://kongchang.com/api/v1/knowledge/claims/900514MCP
get_claim(id=900514)