Unverified50% confidenceConceptExact time
Cross-case contamination refers to the accidental leakage of the previous case's conversation state into a subsequent case in sequentially executed benchmark tests, causing the model to face a contaminated input baseline in the subsequent case.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedData contamination refers to test problems having already appeared in pretraining corpora, causing a model's high scores to more closely reflect memorization rather than reasoning.74% similarUnverifiedBenchmark contamination refers to the situation where benchmark questions have entered a model's training data, causing the model to "recall answers" rather than genuinely reason, inflating benchmark scores and failing to reflect actual capability.68% similarUnverifiedOn contaminated benchmarks, models can directly extract answers from parametric memory rather than relying on in-context retrieval, rendering the reported lossless-compression conclusions meaningless.66% similarUnverified激活修补的操作是对同一模型运行干净输入和污染输入，记录干净运行的中间激活值并注入污染运行，观察哪些替换能恢复正确输出66% similarUnverifiedHot-switching models mid-session in a conversation with existing history causes context contamination, causing the new model to inherit the intermediate state and erroneous assumptions of the old conversation, leading to deviations in the reasoning path.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1007760API
curl https://kongchang.com/api/v1/knowledge/claims/1007760MCP
get_claim(id=1007760)