Author

Yanzeng Li

2 papers indexed here

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Preprint Sep 2026

Are Benchmarks Reliable? Toward Structural Diagnosis via Item-Level Capability Boundaries

Evaluating large language models (LLMs) relies heavily on benchmark scores, yet aggregate metrics can obscure whether individual items serve their intended measurement purpose. BSDProbe provides an item-level framework for benchmark structural diagnosis that jointly characterizes correctness transitions and their suppo...

Hai Hu, Yu-Zhu Liang, Wei Tang et al. · 0 citations

#artificial intelligence Preprint Sep 2026

Are Benchmarks Reliable? Toward Structural Diagnosis via Sample-Level Capability Boundaries

Experiments show that benchmark reliability is axis-conditioned and heterogeneous: GSM8K and MATH exhibit the most stable measurement structures, MMLU and TriviaQA are relatively stable but heterogeneous, while GPQA and PopQA show stronger axis-conditioned risks.

Hai Hu, Yu-Zhu Liang, Wei Tang et al. · 0 citations