Paper
Dados & Embeddings
Zero Gap Is Not Restoration: Stratified Per-Question Probability Evaluation and Step-wise Mitigation of Benchmark Contamination
Proposed SA-PPG metric and RailCap decoding intervention more accurately assess and reduce benchmark contamination in pretrained models.
Hugging Face · Daily Papers
·Ruijie Hou, Yueyang Jiao
·
·▲ 3 upvotes
Este artigo está em destaque na seleção diária de papers do Hugging Face, curada pela comunidade de pesquisa em IA.
Autores: Ruijie Hou, Yueyang Jiao, Zhao Wang, Yingming Li
- 3 upvotes da comunidade
- Temas: contamination mitigation evaluation, G-AP, SA-PPG, per-question probability gaps, RailCap, greedy trajectory
Resumo
Resumo original (em inglês), extraído do paper:
Proposed SA-PPG metric and RailCap decoding intervention more accurately assess and reduce benchmark contamination in pretrained models.