TY  - RPRT
TI  - Revealing what benchmark scores hide: Jointly measuring the breadth and reliability of LLM capabilities using accessibility profiles
AU  - Kazem Faghih
AU  - Yize Cheng
AU  - Shoumik Saha
AU  - Mobina Pournemat
AU  - Armin Gerami
AU  - Soheil Feizi
PY  - 2026
UR  - https://arxiv.org/abs/2610.24957
ID  - 2610.24957
ER  -