# Mellow: a small audio language model for reasoning

arXiv: https://arxiv.org/abs/2503.08540 · HTML: https://mlsanity.com/paper/2503.08540/

## Key facts

- Cited by 27 papers in the map (data through 14 Oct 2026).
- Placed in Audio-language model reasoning, one of 17 directions in LLM reasoning.
- Builds on 24 of its 74 references that are in the map.
- First posted on arXiv 11 Mar 2025.

## Placed in

- LLM reasoning / [Audio-language model reasoning](https://mlsanity.com/lr/audio-language-model-reasoning/)

## Builds on (most cited first, 24 of 24)

- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Gemma 2: Improving Open Language Models at a Practical Size](https://mlsanity.com/paper/2408.00118/) (2024)
- [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://mlsanity.com/paper/2404.14219/) (2024)
- [SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features](https://mlsanity.com/paper/2502.14786/) (2025)
- [Gemma: Open Models Based on Gemini Research and Technology](https://mlsanity.com/paper/2403.08295/) (2024)
- [PaliGemma: A versatile 3B VLM for transfer](https://mlsanity.com/paper/2407.07726/) (2024)
- [Qwen-Audio: Advancing Universal Audio Understanding via Unified Large-Scale Audio-Language Models](https://mlsanity.com/paper/2311.07919/) (2023)
- [AudioLDM: Text-to-Audio Generation with Latent Diffusion Models](https://mlsanity.com/paper/2301.12503/) (2023)
- [Textbooks Are All You Need](https://mlsanity.com/paper/2306.11644/) (2023)
- [Textbooks Are All You Need II: phi-1.5 technical report](https://mlsanity.com/paper/2309.05463/) (2023)
- [SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model](https://mlsanity.com/paper/2502.02737/) (2025)
- [MMAU: A Massive Multi-Task Audio Understanding and Reasoning Benchmark](https://mlsanity.com/paper/2410.19168/) (2024)
- [WavCaps: A ChatGPT-Assisted Weakly-Labelled Audio Captioning Dataset for Audio-Language Multimodal Research](https://mlsanity.com/paper/2303.17395/) (2023)
- [Listen, Think, and Understand](https://mlsanity.com/paper/2305.10790/) (2023)
- [PaliGemma 2: A Family of Versatile VLMs for Transfer](https://mlsanity.com/paper/2412.03555/) (2024)
- [GAMA: A Large Audio-Language Model with Advanced Audio Understanding and Complex Reasoning Abilities](https://mlsanity.com/paper/2406.11768/) (2024)
- [Visual Description Grounding Reduces Hallucinations and Boosts Reasoning in LVLMs](https://mlsanity.com/paper/2405.15683/) (2024)
- [On Metric Learning for Audio-Text Cross-Modal Retrieval](https://mlsanity.com/paper/2203.15537/) (2022)
- [Audio Entailment: Assessing Deductive Reasoning for Audio Understanding](https://mlsanity.com/paper/2407.18062/) (2024)
- [Computer Audition: From Task-Specific Machine Learning to Foundation Models](https://mlsanity.com/paper/2407.15672/) (2024)
- [Vision Language Models Are Few-Shot Audio Spectrogram Classifiers](https://mlsanity.com/paper/2411.12058/) (2024)
- [MACE: Leveraging Audio for Evaluating Audio Captioning Systems](https://mlsanity.com/paper/2411.00321/) (2024)

## Built on by (most cited first, 27 of 27)

- [MMAR: A Challenging Benchmark for Deep Reasoning in Speech, Audio, Music, and Their Mix](https://mlsanity.com/paper/2505.13032/) (2025)
- [Audio-Thinker: Guiding Audio Language Model When and How to Think via Reinforcement Learning](https://mlsanity.com/paper/2508.08039/) (2025)
- [Audio Flamingo Next: Next-Generation Open Audio-Language Models for Speech, Sound, and Music](https://mlsanity.com/paper/2604.10905/) (2026)
- [SpeechR: A Benchmark for Speech Reasoning in Large Audio-Language Models](https://mlsanity.com/paper/2508.02018/) (2025)
- [Evaluating Robustness of Large Audio Language Models to Audio Injection: An Empirical Study](https://mlsanity.com/paper/2505.19598/) (2025)
- [UALM: Unified Audio Language Model for Understanding, Generation and Reasoning](https://mlsanity.com/paper/2510.12000/) (2025)
- [WavBench: Benchmarking Reasoning, Colloquialism, and Paralinguistics for End-to-End Spoken Dialogue Models](https://mlsanity.com/paper/2602.12135/) (2026)
- [Audio Flamingo Sound-CoT Technical Report: Improving Chain-of-Thought Reasoning in Sound Understanding](https://mlsanity.com/paper/2508.11818/) (2025)
- [Audio-Maestro: Enhancing Large Audio-Language Models with Tool-Augmented Reasoning](https://mlsanity.com/paper/2510.11454/) (2025)
- [Spatial Audio Question Answering and Reasoning on Dynamic Source Movements](https://mlsanity.com/paper/2602.16334/) (2026)
- [Empowering Multimodal LLMs with External Tools: A Comprehensive Survey](https://mlsanity.com/paper/2508.10955/) (2025)
- [WoW-Bench: Evaluating Fine-Grained Acoustic Perception in Audio-Language Models via Marine Mammal Vocalizations](https://mlsanity.com/paper/2508.20976/) (2025)
- [CMDAR: A Chinese Multi-scene Dynamic Audio Reasoning Benchmark with Diverse Challenges](https://mlsanity.com/paper/2509.22461/) (2025)
- [SonicBench: Dissecting the Physical Perception Bottleneck in Large Audio Language Models](https://mlsanity.com/paper/2601.11039/) (2026)
- [MUGEN: Evaluating and Improving Multi-audio Understanding of Large Audio-Language Models](https://mlsanity.com/paper/2603.09714/) (2026)
- [Heard but Not Heeded: Paralinguistic Information Encoding and Loss in Audio-Language Models](https://mlsanity.com/paper/2609.00727/) (2026)
- [Enhancing Audio Reasoning via Semantic Summary Prediction](https://mlsanity.com/paper/2609.20849/) (2026)
- [SAM: A Mamba-2 State-Space Audio-Language Model](https://mlsanity.com/paper/2509.15680/) (2025)
- [AURA Score: A Metric For Holistic Audio Question Answering Evaluation](https://mlsanity.com/paper/2510.04934/) (2025)
- [BASS: Benchmarking Audio LMs for Musical Structure and Semantic Reasoning](https://mlsanity.com/paper/2602.04085/) (2026)
- [AudioDER: A Deduplication-Enhanced Reasoning Dataset for Post-Training Large Audio-Language Models](https://mlsanity.com/paper/2606.14591/) (2026)
- [Noisy-Channel Minimum Bayes Risk Decoding](https://mlsanity.com/paper/2607.05198/) (2026)
- [Training-Free Speech-Centric Omni Understanding with Frozen VLMs](https://mlsanity.com/paper/2609.04242/) (2026)
- [Knowing When Not to Answer: Pseudo-Ensembles for Abstention in Music Audio-Language Models](https://mlsanity.com/paper/2609.04362/) (2026)
- [Samsone: A Family of Open Small Audio Language Models for On-Device Inference](https://mlsanity.com/paper/2609.21666/) (2026)
- [ParA-LLM: A Unified Approach to Paralinguistic and Acoustic Speech Understanding](https://mlsanity.com/paper/2609.22771/) (2026)
- [Mizar: A 159M-Parameter Audio-Language Model for Audio Understanding](https://mlsanity.com/paper/2609.28344/) (2026)

## Abstract (arXiv)

Multimodal Audio-Language Models (ALMs) can understand and reason over both audio and text. Typically, reasoning performance correlates with model size, with the best results achieved by models exceeding 8 billion parameters. However, no prior work has explored enabling small audio-language models to perform reasoning tasks, despite the potential applications for edge devices. To address this gap, we introduce Mellow, a small Audio-Language Model specifically designed for reasoning. Mellow achieves state-of-the-art performance among existing small audio-language models and surpasses several larger models in reasoning capabilities. For instance, Mellow scores 52.11 on MMAU, comparable to SoTA Qwen2 Audio (which scores 52.5) while using 50 times fewer parameters and being trained on 60 times less data (audio hrs). To train Mellow, we introduce ReasonAQA, a dataset designed to enhance audio-grounded reasoning in models. It consists of a mixture of existing datasets (30% of the data) and synthetically generated data (70%). The synthetic dataset is derived from audio captioning datasets, where Large Language Models (LLMs) generate detailed and multiple-choice questions focusing on audio events, objects, acoustic scenes, signal properties, semantics, and listener emotions. To evaluate Mellow's reasoning ability, we benchmark it on a diverse set of tasks, assessing on both in-distribution and out-of-distribution data, including audio understanding, deductive reasoning, and comparative reasoning. Finally, we conduct extensive ablation studies to explore the impact of projection layer choices, synthetic data generation methods, and language model pretraining on reasoning performance. Our training dataset, findings, and baseline pave the way for developing small ALMs capable of reasoning.