PulseAugur
EN
LIVE 09:49:49

New DEAF benchmark reveals Audio MLLMs rely on text over sound

By PulseAugur Editorial · [1 sources] ·

Researchers have introduced DEAF, a new benchmark designed to evaluate the acoustic faithfulness of Audio Multimodal Large Language Models (Audio MLLMs). The benchmark includes over 2,700 stimuli across emotional prosody, background sounds, and speaker identity to test whether these models truly process acoustic signals or rely on textual inference. Evaluations of seven Audio MLLMs showed a consistent tendency for models to be driven by textual inputs, indicating a gap between their performance on standard speech benchmarks and genuine acoustic understanding. AI

IMPACT Highlights a critical limitation in current Audio MLLMs, potentially guiding future research towards models with genuine acoustic understanding.

RANK_REASON The cluster contains an academic paper introducing a new benchmark for evaluating AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

New DEAF benchmark reveals Audio MLLMs rely on text over sound

How we ranked this

Signal score
12 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The cluster contains an academic paper introducing a new benchmark for evaluating AI models. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. arXiv cs.AI TIER_1 English(EN) · Jiaqi Xiong, Yunjia Qi, Qi Cao, Yu Zheng, Yutong Zhang, Ziteng Wang, Ruofan Liao, Weisheng Xu, Sichen Liu ·

    DEAF: A Benchmark for Diagnostic Evaluation of Acoustic Faithfulness in Audio Language Models

    arXiv:2603.18048v3 Announce Type: replace Abstract: Recent Audio Multimodal Large Language Models (Audio MLLMs) demonstrate impressive performance on speech benchmarks, yet it remains unclear whether these models genuinely process acoustic signals or rely on text-based semantic i…

RELATED ENTITIES

RELATED TOPICS