AU-Harness: An Open-Source Toolkit for Holistic Evaluation of AU dio-LLMs
Sidharth Surapaneni, Hoang Nguyen, Jash Mehta, Aman Tiwari,, Oluwanifemi Bamgbose, Akshay Kalkunte, Sai Rajeswar, Sathwik Tejaswi Madhusudhan · 2025
Read AU-Harness: An Open-Source Toolkit for Holistic Evaluation of AU dio-LLMs by Sidharth Surapaneni, Hoang Nguyen, Jash Mehta, Aman Tiwari,, Oluwanifemi Bamgbose, Akshay Kalkunte, Sai Rajeswar, Sathwik Tejaswi Madhusudhan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-08031.txt.
Description
Audio language models finally get a shared measuring stick in a 2025 toolkit paper by Hoang Nguyen and 11 co-authors. Large audio models are advancing quickly, but evaluation tools are slow, non-standard, and hard to compare across studies. Long multi-turn conversations, where context must carry across exchanges, are especially poorly supported by existing frameworks.
AU-Harness is open-source and built for holistic evaluation, aiming at faster pipelines and unified, scalable assessment that keeps pace with model growth. Fair comparison is not bureaucratic tidiness: without it, claims about listening, reasoning, and dialogue ability cannot be checked. A common harness lets researchers see which audio models genuinely improve and which merely test well on friendly setups.
Catalog record
| CREATOR | Sidharth Surapaneni, Hoang Nguyen, Jash Mehta, Aman Tiwari,, Oluwanifemi Bamgbose, Akshay Kalkunte, Sai Rajeswar, Sathwik Tejaswi Madhusudhan |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-08031