AI Navigate

ALARM: Audio-Language Alignment for Reasoning Models

arXiv cs.CL / 3/11/2026

Signals & Early TrendsIdeas & Deep AnalysisModels & Research

Key Points

  • The paper introduces ALARM, a large audio language model that extends large language models (LLMs) with enhanced auditory understanding, focusing on reasoning capabilities.
  • It addresses challenges in training reasoning LLMs (RLMs) by proposing self-rephrasing to generate audio-understanding responses aligned with RLM reasoning processes.
  • ALARM incorporates fusion and compression of multiple audio encoders to create stronger audio feature representations.
  • The model is trained on a large, diverse 6-million-instance multi-task corpus with 19,000 hours of speech, music, and sound, resulting in a 4B-parameter model.
  • ALARM achieves superior performance compared to similarly sized models and ranks third overall on benchmarks like MMAU-speech and MMSU, also maintaining low training costs and preserving text capabilities.

Computer Science > Computation and Language

arXiv:2603.09556 (cs)
[Submitted on 10 Mar 2026]

Title:ALARM: Audio-Language Alignment for Reasoning Models

View a PDF of the paper titled ALARM: Audio-Language Alignment for Reasoning Models, by Petr Grinberg and 1 other authors
View PDF HTML (experimental)
Abstract:Large audio language models (ALMs) extend LLMs with auditory understanding. A common approach freezes the LLM and trains only an adapter on self-generated targets. However, this fails for reasoning LLMs (RLMs) whose built-in chain-of-thought traces expose the textual surrogate input, yielding unnatural responses. We propose self-rephrasing, converting self-generated responses into audio-understanding variants compatible with RLMs while preserving distributional alignment. We further fuse and compress multiple audio encoders for stronger representations. For training, we construct a 6M-instance multi-task corpus (2.5M unique prompts) spanning 19K hours of speech, music, and sound. Our 4B-parameter ALM outperforms similarly sized models and surpasses most larger ALMs on related audio-reasoning benchmarks, while preserving textual capabilities with a low training cost. Notably, we achieve the best open-source result on the MMAU-speech and MMSU benchmarks and rank third among all the models.
Comments:
Subjects: Computation and Language (cs.CL)
Cite as: arXiv:2603.09556 [cs.CL]
  (or arXiv:2603.09556v1 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2603.09556
Focus to learn more
arXiv-issued DOI via DataCite

Submission history

From: Hassan Shahmohammadi [view email]
[v1] Tue, 10 Mar 2026 12:03:25 UTC (1,973 KB)
Full-text links:

Access Paper:

Current browse context:
cs.CL
< prev   |   next >
Change to browse by:
cs

References & Citations

export BibTeX citation Loading...

BibTeX formatted citation

×
Data provided by:

Bookmark

BibSonomy logo Reddit logo
Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
Links to Code Toggle
Papers with Code (What is Papers with Code?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos

Demos

Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers

Recommenders and Search Tools

Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.