# WoW-Bench: Evaluating Fine-Grained Acoustic Perception in Audio-Language Models via Marine Mammal Vocalizations

## TL;DR

This paper introduces WoW-Bench, a new benchmark for evaluating the low-level auditory perception of large audio language models using marine mammal vocalizations, revealing current models' limitations compared to humans.

## Contribution

The paper presents WoW-Bench, a novel evaluation framework for assessing fine-grained acoustic perception in audio-language models, focusing on low-level listening skills.

## Key findings

- State-of-the-art LALMs perform far below human levels.
- WoW-Bench effectively differentiates model capabilities in acoustic perception.
- Models struggle with tasks involving distractors, indicating reliance on heuristics.

## Abstract

Large audio language models (LALMs) extend language understanding into the auditory domain, yet their ability to perform low-level listening, such as pitch and duration detection, remains underexplored. However, low-level listening is critical for real-world, out-of-distribution tasks where models must reason about unfamiliar sounds based on fine-grained acoustic cues. To address this gap, we introduce the World-of-Whale benchmark (WoW-Bench) to evaluate low-level auditory perception and cognition using marine mammal vocalizations. WoW-bench is composed of a Perception benchmark for categorizing novel sounds and a Cognition benchmark, inspired by Bloom's taxonomy, to assess the abilities to remember, understand, apply, and analyze sound events. For the Cognition benchmark, we additionally introduce distractor questions to evaluate whether models are truly solving problems through listening rather than relying on other heuristics. Experiments with state-of-the-art LALMs show performance far below human levels, indicating a need for stronger auditory grounding in LALMs.