# Do Audio LLMs Listen or Read? Analyzing and Mitigating Paralinguistic Failures with VoxParadox

arXiv: https://arxiv.org/abs/2605.27772 · HTML: https://mlsanity.com/paper/2605.27772/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 40 of its 57 references that are in the map.
- First posted on arXiv 26 May 2026.

## Builds on (most cited first, 30 of 40)

- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://mlsanity.com/paper/2106.07447/) (2021)
- [VoxCeleb2: Deep Speaker Recognition](https://mlsanity.com/paper/1806.05622/) (2018)
- [SUPERB: Speech processing Universal PERformance Benchmark](https://mlsanity.com/paper/2105.01051/) (2021)
- [SpeechGPT: Empowering Large Language Models with Intrinsic Cross-Modal Conversational Abilities](https://mlsanity.com/paper/2305.11000/) (2023)
- [Perception Encoder: The best visual embeddings are not at the output of the network](https://mlsanity.com/paper/2504.13181/) (2025)
- [Layer-wise Analysis of a Self-supervised Speech Representation Model](https://mlsanity.com/paper/2107.04734/) (2021)
- [MMAU: A Massive Multi-Task Audio Understanding and Reasoning Benchmark](https://mlsanity.com/paper/2410.19168/) (2024)
- [Audio Flamingo 3: Advancing Audio Intelligence with Fully Open Large Audio Language Models](https://mlsanity.com/paper/2507.08128/) (2025)
- [Audio Flamingo 2: An Audio-Language Model with Long-Audio Understanding and Expert Reasoning Abilities](https://mlsanity.com/paper/2503.03983/) (2025)
- [MMAR: A Challenging Benchmark for Deep Reasoning in Speech, Audio, Music, and Their Mix](https://mlsanity.com/paper/2505.13032/) (2025)
- [MMSU: A Massive Multi-task Spoken Language Understanding and Reasoning Benchmark](https://mlsanity.com/paper/2506.04779/) (2025)
- [FOIL it! Find One mismatch between Image and Language caption](https://mlsanity.com/paper/1705.01359/) (2017)
- [Whisper-AT: Noise-Robust Automatic Speech Recognizers are Also Strong General Audio Event Taggers](https://mlsanity.com/paper/2307.03183/) (2023)
- [SD-Eval: A Benchmark Dataset for Spoken Dialogue Understanding Beyond Words](https://mlsanity.com/paper/2406.13340/) (2024)
- [The Emotional Voices Database: Towards Controlling the Emotion Dimension in Voice Generation Systems](https://mlsanity.com/paper/1806.09514/) (2018)
- [MMAU-Pro: A Challenging and Comprehensive Benchmark for Holistic Evaluation of Audio General Intelligence](https://mlsanity.com/paper/2508.13992/) (2025)
- [Open-Source Conversational AI with SpeechBrain 1.0](https://mlsanity.com/paper/2407.00463/) (2024)
- [The OMG-Emotion Behavior Dataset](https://mlsanity.com/paper/1803.05434/) (2018)
- [Do Audio LLMs Really LISTEN, or Just Transcribe? Measuring Lexical vs. Acoustic Emotion Cues Reliance](https://mlsanity.com/paper/2510.10444/) (2025)
- [MMFuser: Multimodal Multi-Layer Feature Fuser for Fine-Grained Vision-Language Understanding](https://mlsanity.com/paper/2410.11829/) (2024)
- [Multi-Layer Visual Feature Fusion in Multimodal LLMs: Methods, Analysis, and Best Practices](https://mlsanity.com/paper/2503.06063/) (2025)
- [VoxCeleb Enrichment for Age and Gender Recognition](https://mlsanity.com/paper/2109.13510/) (2021)
- [SpeechCraft: A Fine-grained Expressive Speech Dataset with Natural Language Description](https://mlsanity.com/paper/2408.13608/) (2024)
- [Frozen Large Language Models Can Perceive Paralinguistic Aspects of Speech](https://mlsanity.com/paper/2410.01162/) (2024)
- [Advancing Zero-shot Text-to-Speech Intelligibility across Diverse Domains via Preference Alignment](https://mlsanity.com/paper/2505.04113/) (2025)
- [Attentive Merging of Hidden Embeddings from Pre-trained Speech Model for Anti-spoofing Detection](https://mlsanity.com/paper/2406.10283/) (2024)
- [Just ASR + LLM? A Study on Speech Large Language Models' Ability to Identify and Understand Speaker in Spoken Dialogue](https://mlsanity.com/paper/2409.04927/) (2024)

## Built on by (most cited first, 7 of 7)

- [Real-Time Voice AI Hears but Does Not Listen](https://mlsanity.com/paper/2606.26083/) (2026)
- [Auditing Protocol-Level Shortcuts in Large Audio Language Model Judges for Speech Evaluation](https://mlsanity.com/paper/2607.13477/) (2026)
- [Do SpeechLMs Hear Their Own Opinions? Diagnosing and Mitigating Previous-Belief Contamination in Streaming Emotion Understanding](https://mlsanity.com/paper/2608.20769/) (2026)
- [Encoder-Side Neuron Identification and Amplification for Acoustic Perception in Large Audio-Language Models](https://mlsanity.com/paper/2607.11801/) (2026)
- [Where Should Language Sit in a Multimodal Model? Lessons from What Language Does to Human Perception and Cognition](https://mlsanity.com/paper/2609.07474/) (2026)
- [Do Audio LLMs Listen Before They Act? Diagnosing Acoustic-Context Gating in Voice Agents](https://mlsanity.com/paper/2609.32536/) (2026)
- [Almost Human, Except When It Matters: VoxParity and the Decisions a Voice Should Change](https://mlsanity.com/paper/2609.35922/) (2026)

## Abstract (arXiv)

Audio large language models (Audio LLMs) demonstrate strong performance on speech understanding tasks, yet their ability to understand paralinguistic information remains limited. To systematically quantify this issue, we introduce VoxParadox, an adversarial benchmark with 2,000 verified examples, spanning 10 paralinguistic tasks, created with controlled speech synthesis to intentionally mismatch transcript claims and speaking style, enabling direct measurement of speech paralinguistic understanding. Evaluation of a diverse set of Audio LLMs reveals consistently low accuracy on acoustic ground truth and a strong tendency to follow language-implied (incorrect) answers. To understand the cause of this gap, we perform layer-wise probing and find that (i) paralinguistic cues can degrade in deeper encoder layers and at the encoder--LLM interface, and (ii) even when such cues are available in audio tokens, the language model frequently ignores them. To address these problems, we propose Prompt-Conditioned Layer Mixer (PCLM), which adaptively combines information from multiple audio layers based on the input prompt, and pair it with Direct Preference Optimization (DPO) to explicitly prefer acoustically supported options over language-implied alternatives. These methods substantially improve Audio LLM paralinguistic understanding, improving Audio Flamingo 3 from 17.40% to 65.20% on VoxParadox, and from 37.74% to 54.78% on MMSU paralinguistic subset. Our project page is available at https://voxparadox.github.io/.