Deconstructing instruction-following: A new benchmark for granular analysis of Large Language Model instruction compliance abilities

A modular framework that uses a dynamically generated dataset to evaluate the capability of Large Language Models.

EACL
March 24, 2026

Latest publications