Elamur Research⌕
☰
home / papers / 2025 / VoxEval: Benchmarking the Knowledge Understanding Capabilities of End-to-End Spoken Language Models

VoxEval: Benchmarking the Knowledge Understanding Capabilities of End-to-End Spoken Language Models

Paper · 2025 · Wenqian Cui, Xiaoqi Jiao, Ziqiao Meng et al.

Summary

Spoken Language Models (SLMs) that both listen and reply in speech need broad world knowledge. Existing audio QA benchmarks still pair audio with text questions or score text answers, so they cannot test true speech-to-speech use or how stable a model is when the same question is spoken differently. The authors built VoxEval by converting the MMLU test set into audio with the OpenAI TTS API. Both questions and answers are in speech, and model replies are transcribed with ASR before accuracy is scored. Each item is rendered under many input audio conditions: different speakers, speaking styles and linguistic variations, audio quality, and pitch. Math problems were rewritten into spoken form, and spoken Chain-of-Thought examples were added to probe step-by-step reasoning. When tested on VoxEval, current SLMs often score below random guessing. Their answers change across audio conditions, and they show weak mathematical reasoning. The benchmark gives developers a speech-only target for improving knowledge, robustness and reasoning.

Key points

  • Scale: VoxEval contains 13,938 QA pairs across 56 MMLU subjects, with 153,318 audio variations in total. The authors present it as the first fully end-to-end SpeechQA benchmark, compared against VoiceBench, MMAU, AudioBench and others.
  • Audio conditions:
    • Voices come from all six OpenAI TTS speakers, with "alloy" as the reference voice.
    • Linguistic variations include filler words, mispronunciations, disfluencies, false starts and non-native grammar errors. These barely affect scores.
    • Pitch shifts are the hardest condition for most models.
  • Robustness:
    • Overall score shifts are usually only 1–2%, but gaps for individual subjects often exceed 10%.
    • TWIST and Moshi are the least stable models, while GLM-4-Voice is the most stable.
  • Math setup: Elementary, high school and college math are tested in two ways:
    • 5-shot direct answering.
    • Spoken CoT using three dev-set examples, with input either trimmed to the last 80 seconds ("Cut") or left whole ("No cut").
    • Answers are extracted by string matching or by GPT-4o.
  • Math results:
    • CoT lowered accuracy for every tested model.
    • Scores barely depend on difficulty level, and only GLM-4-Voice performs reliably on elementary math.
    • The authors acknowledge a limitation: relying on synthetic TTS speech misses natural speech traits such as dialects and emotion.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓National University of SingaporeSingapore · Singapore180 works · 34,304 citations🎓Chinese University of Hong KongHong Kong · China126 works · 14,073 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

Cite

Cite as Wenqian Cui, Xiaoqi Jiao, Ziqiao Meng and Irwin King. 2025. VoxEval: Benchmarking the Knowledge Understanding Capabilities of End-to-End Spoken Language Models. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2025.acl-long.818

@inproceedings{cui2025voxeval,
  title = {{VoxEval: Benchmarking the Knowledge Understanding Capabilities of End-to-End Spoken Language Models}},
  author = {Wenqian Cui, Xiaoqi Jiao, Ziqiao Meng and Irwin King},
  booktitle = {Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2025},
  doi = {10.18653/v1/2025.acl-long.818},
  url = {https://doi.org/10.18653/v1/2025.acl-long.818}
}

References

38 references, 24 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jinze Bai, Shuai Bai, Yunfei Chu, Zeyu Cui, Kai Dang, Xiaodong Deng, Yang Fan, Wenbin Ge, Yu Han, Fei Huang, et al. 2023. Qwen technical report. arXiv preprint arXiv:2309.16609.
  2. Qian Chen, Yafeng Chen, Yanni Chen, Mengzhe Chen, Yingda Chen, Chong Deng, Zhihao Du, Ruize Gao, Changfeng Gao, Zhifu Gao, et al. 2025. Minmo: A multimodal large language model for seamless voice interaction. arXiv preprint arXiv:2501.06282. — [arXiv:2501.06282](https://arxiv.org/abs/2501.06282) find
  3. Sanyuan Chen, Yu Wu, Chengyi Wang, Shujie Liu, Daniel Tompkins, Zhuo Chen, Wanxiang Che, Xiangzhan Yu, and Furu Wei. 2023. Beats: audio pre-training with acoustic tokenizers. In Proceedings of the 40th International Conference on Machine Learning, ICML’23. JMLR.org. find
  4. Yiming Chen, Xianghu Yue, Chen Zhang, Xiaoxue Gao, Robby T Tan, and Haizhou Li. 2024. Voicebench: Benchmarking llm-based voice assistants. arXiv preprint arXiv:2410.17196.
  5. Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E. Gonzalez, Ion Stoica, and Eric P. Xing. 2023. Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality. find
  6. Yunfei Chu, Jin Xu, Xiaohuan Zhou, Qian Yang, Shiliang Zhang, Zhijie Yan, Chang Zhou, and Jingren Zhou. 2023. Qwen-audio: Advancing universal audio understanding via unified large-scale audio-language models. arXiv preprint arXiv:2311.07919.
  7. Yu-An Chung, Yu Zhang, Wei Han, Chung-Cheng Chiu, James Qin, Ruoming Pang, and Yonghui Wu. 2021. W2v-bert: Combining contrastive learning and masked language modeling for self-supervised speech pre-training. In 2021 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), pages 244–250.… find
  8. Wenqian Cui, Dianzhi Yu, Xiaoqi Jiao, Ziqiao Meng, Guangyan Zhang, Qichao Wang, Yiwen Guo, and Irwin King. 2024. Recent advances in speech language models: A survey. arXiv preprint arXiv:2410.03751.
  9. Alexandre Défossez, Laurent Mazaré, Manu Orsini, Amélie Royer, Patrick Pérez, Hervé Jégou, Edouard Grave, and Neil Zeghidour. 2024. Moshi: a speech-text foundation model for real-time dialogue. Technical report, Kyutai.
  10. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  11. Michael Hassid, Tal Remez, Tu Anh Nguyen, Itai Gat, Alexis Conneau, Felix Kreuk, Jade Copet, Alexandre Defossez, Gabriel Synnaeve, Emmanuel Dupoux, et al. 2024. Textually pretrained speech language models. Advances in Neural Information Processing Systems, 36. find
  12. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021. Measuring massive multitask language understanding. In Proceedings of the 9th International Conference on Learning Representations (ICLR).
  13. Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, and Abdelrahman Mohamed. 2021. Hubert: Self-supervised speech representation learning by masked prediction of hidden units. IEEE/ACM transactions on audio, speech, and language processing, 29:3451–3460.
  14. Chien-yu Huang, Wei-Chih Chen, Shu-wen Yang, Andy T Liu, Chen-An Li, Yu-Xiang Lin, Wei-Cheng Tseng, Anuj Diwan, Yi-Jen Shih, Jiatong Shi, et al. 2024. Dynamic-superb phase-2: A collaboratively expanding benchmark for measuring the capabilities of spoken language models with 180 tasks. arXiv preprint… — [arXiv:2411.05361](https://arxiv.org/abs/2411.05361) find
  15. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, et al. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276.
  16. Eugene Kharitonov, Ann Lee, Adam Polyak, Yossi Adi, Jade Copet, Kushal Lakhotia, Tu Anh Nguyen, Morgane Riviere, Abdelrahman Mohamed, Emmanuel Dupoux, and Wei-Ning Hsu. 2022. Text-free prosody-aware generative spoken language modeling. In Proceedings of the 60th Annual Meeting of the Association for…
  17. Jungil Kong, Jaehyeon Kim, and Jaekyoung Bae. 2020. Hifi-gan: Generative adversarial networks for efficient and high fidelity speech synthesis. Advances in neural information processing systems, 33:17022– 17033. find
  18. Zuwei Long, Yunhang Shen, Chaoyou Fu, Heting Gao, Lijiang Li, Peixian Chen, Mengdan Zhang, Hang Shao, Jian Li, Jinlong Peng, et al. 2025. Vita-audio: Fast interleaved cross-modal token generation for efficient large speech-language model. arXiv preprint arXiv:2505.03739. — [arXiv:2505.03739](https://arxiv.org/abs/2505.03739) find
  19. Todor Mihaylov, Peter Clark, Tushar Khot, and Ashish Sabharwal. 2018. Can a suit of armor conduct electricity? a new dataset for open book question answering. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 2381–2391, Brussels, Belgium. Association fo… find
  20. Tu Anh Nguyen, Maureen de Seyssel, Patricia Rozé, Morgane Rivière, Evgeny Kharitonov, Alexei Baevski, Ewan Dunbar, and Emmanuel Dupoux. 2020. The zero resource speech benchmark 2021: Metrics and baselines for unsupervised spoken language modeling. In Proceedings of the Workshop on Self-Supervised Le… find
  21. Tu Anh Nguyen, Benjamin Muller, Bokai Yu, Marta R Costa-Jussa, Maha Elbayad, Sravya Popuri, Paul-Ambroise Duquenne, Robin Algayres, Ruslan Mavlyutov, Itai Gat, et al. 2024. Spirit-lm: Interleaved spoken and written language model. arXiv preprint arXiv:2402.05755.
  22. Adam Polyak, Yossi Adi, Jade Copet, Eugene Kharitonov, Kushal Lakhotia, Wei-Ning Hsu, Abdelrahman Mohamed, and Emmanuel Dupoux. 2021. Speech resynthesis from discrete disentangled self-supervised representations. In Proc. find
  23. Interspeech 2021, pages 3615–3619. find
  24. Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2023. Robust speech recognition via large-scale weak supervision. In International Conference on Machine Learning, pages 28492–28518. PMLR.
  25. S Sakshi, Utkarsh Tyagi, Sonal Kumar, Ashish Seth, Ramaneswaran Selvakumar, Oriol Nieto, Ramani Duraiswami, Sreyan Ghosh, and Dinesh Manocha. 2024. Mmau: A massive multi-task audio understanding and reasoning benchmark. arXiv preprint arXiv:2410.19168. find
  26. Alon Talmor, Jonathan Herzig, Nicholas Lourie, and Jonathan Berant. 2019. CommonsenseQA: A question answering challenge targeting commonsense knowledge. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,…
  27. Changli Tang, Wenyi Yu, Guangzhi Sun, Xianzhao Chen, Tian Tan, Wei Li, Lu Lu, Zejun Ma, and Chao Zhang. 2024. Salmonn: Towards generic hearing abilities for large language models. In Proceedings of the 12th International Conference on Learning Representations (ICLR).
  28. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. 2023. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288.
  29. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Ł ukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc.
  30. Bin Wang, Xunlong Zou, Geyu Lin, Shuo Sun, Zhuohan Liu, Wenyu Zhang, Zhengyuan Liu, AiTi Aw, and Nancy F Chen. 2024a. Audiobench: A universal benchmark for audio large language models. arXiv preprint arXiv:2406.16020.
  31. Yubo Wang, Xueguang Ma, Ge Zhang, Yuansheng Ni, Abhranil Chandra, Shiguang Guo, Weiming Ren, Aaran Arulraj, Xuan He, Ziyan Jiang, et al. 2024b. Mmlu-pro: A more robust and challenging multi-task language understanding benchmark. arXiv preprint arXiv:2406.01574.
  32. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V. Le, and Denny Zhou. 2022. Chain-of-thought prompting elicits reasoning in large language models. In Proceedings of the 36th International Conference on Neural Information Processing Systems, NIPS ’22, R…
  33. Qian Yang, Jin Xu, Wenrui Liu, Yunfei Chu, Ziyue Jiang, Xiaohuan Zhou, Yichong Leng, Yuanjun Lv, Zhou Zhao, Chang Zhou, et al. 2024. Air-bench: Benchmarking large audio-language models via generative comprehension. arXiv preprint arXiv:2402.07729.
  34. Neil Zeghidour, Alejandro Luebs, Ahmed Omran, Jan Skoglund, and Marco Tagliasacchi. 2021. Soundstream: An end-to-end neural audio codec. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 30:495–507.
  35. Aohan Zeng, Zhengxiao Du, Mingdao Liu, Kedong Wang, Shengmin Jiang, Lei Zhao, Yuxiao Dong, and Jie Tang. 2024. Glm-4-voice: Towards intelligent and human-like end-to-end spoken chatbot. arXiv preprint arXiv:2412.02612.
  36. Dong Zhang, Shimin Li, Xin Zhang, Jun Zhan, Pengyu Wang, Yaqian Zhou, and Xipeng Qiu. 2023. SpeechGPT: Empowering large language models with intrinsic cross-modal conversational abilities. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 15757–15773, Singapore. Associa…
  37. Xin Zhang, Dong Zhang, Shimin Li, Yaqian Zhou, and Xipeng Qiu. 2024. Speechtokenizer: Unified speech WER CER Total Word Errors Total Character Errors Total Words in GT Total Characters in GT 0.01 0.01 3 13 300 1759 0.16 0.10 12 22 78 401 0.08 0.04 23 64 308 1767 *Table 9: ASR process evaluation resu… find
  38. Wanjun Zhong, Ruixiang Cui, Yiduo Guo, Yaobo Liang, Shuai Lu, Yanlin Wang, Amin Saied, Weizhu Chen, and Nan Duan. 2024. AGIEval: A human-centric benchmark for evaluating foundation models. In Findings of the Association for Computational Linguistics: NAACL 2024, pages 2299–2314, Mexico City, Mexico.…

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/cui-2025-voxeval-benchmarking-knowledge-understanding-capabilities/