BLAB (Brutally Long Audio Bench) is a dataset designed to evaluate the multimodal understanding abilities of audio language models (LMs). It focuses on audio perception and reasoning, with eight distinct audio tasks across four categories: localization, counting, emotion, and duration estimation. It assesses audio LMs that accept both text and audio as input and generate text.
No results indexed yet — be the first to submit a score.
Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.