Unverified80% confidenceFactExact time
Multiple studies in 2023 found that abnormally high scores from certain models on specific benchmarks were highly correlated with data leakage
1
Sources
80%
Confidence
Long-term
Relevance
8/3/2026
First Seen
Sources
Text Arena: A Deep Dive into the LLM Battle Evaluation Platform and Its Ranking Mechanism
Related Entities
Related Claims
Unverified基准测试面临饱和（Saturation）和数据污染（Data Contamination）两个系统性挑战，当多个主流模型得分超过95%时测试集失去区分能力67% similarUnverifiedSince the second half of 2024, the industry has broadly observed that the improvement margins of mainstream large models on standard evaluation benchmarks have narrowed significantly, a phenomenon referred to as 'Scaling Law deceleration.'63% similarUnverifiedMultiple independent studies in 2023 showed that some top models scored extremely high on standard mathematical reasoning tests but performed significantly worse on equivalent rephrased versions, indicating that the models learned test patterns rather than genuine reasoning ability.63% similarUnverifiedAI benchmark suites like MMLU, HumanEval, and MATH may produce inflated scores due to benchmark contamination, where models have indirectly encountered test data during training.62% similarUnverifiedWhen samples are severely imbalanced, Accuracy becomes misleading—a model that predicts everyone as healthy can still achieve an accuracy of over 99%.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/680095API
curl https://kongchang.com/api/v1/knowledge/claims/680095MCP
get_claim(id=680095)