Unverified50% confidenceFactExact time
Benchmark gaming refers to the phenomenon where developers optimize targeted for specific benchmarks (such as adding training data stylistically similar to target benchmark questions during the fine-tuning phase) to inflate scores without improving genuine general capabilities
1
Sources
50%
Confidence
Long-term
Relevance
9/8/2026
First Seen
Sources
Related Entities
Related Claims
Unverified评测游戏（Benchmark Gaming）指AI公司通过选择性使用特定基准测试来最大化模型表现分数的行为，包括数据污染和针对特定评测集过度优化77% similarUnverifiedBenchmark hacking refers to models potentially being optimized for specific test sets rather than possessing general capabilities, resulting in a gap between benchmark scores and real-world performance.71% similarUnverified基准测试本身就是一种数据激活的手段，为领域数据赋予清晰的优化方向71% similarUnverifiedFrom checkpoint testing to official release, a model typically still needs to go through processes such as fine-tuning optimization, safety hardening, inference efficiency optimization, and API adaptation.71% similarUnverifiedDevelopers should establish benchmarks with fixed test cases to regularly evaluate the output quality of platform models, rather than judging by intuition.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/875818API
curl https://kongchang.com/api/v1/knowledge/claims/875818MCP
get_claim(id=875818)