Robust Summarization of Doctor-Patient Conversations: TalTech Systems for the Beyond Transcription Challenge

Aivo Olev; Tanel Alum\"ae

REVIEW 3 major objections 5 minor 12 references

Reinforcement learning against a medical-concept metric can improve direct audio-to-SOAP summarization without increasing hallucination, the top result in both tracks of the challenge.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 18:37 UTC pith:T6CLXUMQ

load-bearing objection Winning BeTraC systems with a practical RL recipe, but the core safety claim leans on an external LLM judge nobody has validated. the 3 major comments →

arxiv 2607.17230 v1 pith:T6CLXUMQ submitted 2026-07-19 cs.CL eess.AS

Robust Summarization of Doctor-Patient Conversations: TalTech Systems for the Beyond Transcription Challenge

Aivo Olev , Tanel Alum\"ae This is my paper

classification cs.CL eess.AS

keywords direct speech summarizationSOAP note generationOpen Medical Concept F1reinforcement learninghallucination evaluationdoctor-patient conversationspeech LLM fine-tuningdomain robustness

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tries to establish that a speech LLM can be trained to produce clinical SOAP notes directly from long doctor-patient audio, skipping any transcription step, and that the training recipe does not trade factual safety for metric score. The authors fine-tune open-weight speech models with supervised learning on reference notes and then reinforcement learning whose reward is the challenge's Open Medical Concept F1. Their systems placed first in both lightweight and heavyweight tracks, and an independent judge-based evaluation found the lowest hallucination rates in both tracks. The conclusion is that optimizing a concept-matching metric through reinforcement learning need not cause the expected degeneracy of listing many plausible medical concepts, because the precision term in F1 and length discipline keep the notes concise.

Core claim

Starting from zero-shot screening that identified models able to process roughly nine-minute conversations, the authors adapted a small and a large open-weight speech LLM using low-rank supervised fine-tuning followed by reinforcement learning with Open Medical Concept F1 as the reward. On the blind test set the adapted systems achieved Concept F1 of 0.563 (heavyweight) and 0.543 (lightweight), both more than double the baseline, and an LLM-as-judge evaluation measured hallucination rates of 0.08% and 0.25% on the main test split—the lowest in each track. The paper takes this as evidence that RL against such a metric can improve clinical summarization without inflating note length or encoura

What carries the argument

Open Medical Concept F1—a metric that scores generated notes by MeSH keyword matching and medical named-entity recognition against reference notes—is the load-bearing object: it serves both as training reward and ranking metric. Around it, the recipe couples low-rank supervised fine-tuning (to teach note format) with an on-policy reinforcement learning update that maximizes the metric on rollouts while keeping notes at roughly 295 words. The paper also relies on a zero-shot screening stage that used word-error rate on long audio as a diagnostic for whether a model can follow the full conversation, and on an LLM-based judge that decomposes each note into atomic claims to measure faithfulness

Load-bearing premise

The paper's central safety claim—that reinforcement learning does not increase hallucinations—rests on an LLM judge that breaks each note into atomic claims and checks them against the transcript, with no human agreement study, error analysis, or confidence intervals reported.

What would settle it

Take the systems' generated notes, inject known hallucinated symptoms or diagnoses into a sample, and have independent clinician annotators mark factual errors; if the human error rate is materially higher than the LLM judge's 0.08–0.25% hallucination rates, the conclusion that the RL recipe preserves factual reliability fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

Reinforcement learning against Open Medical Concept F1 adds 0.04–0.05 Concept F1 over supervised fine-tuning alone, without lengthening notes or increasing judged hallucinations.
Fine-tuning on text transcripts transfers to speech input with only a small drop, so audio-input fine-tuning can be skipped for large models when computing budget is tight.
Models fine-tuned on normalized transcripts generalize better from synthetic to acted and real audio, losing only 0.008 Concept F1 versus 0.038 for the audio-fine-tuned lightweight system.
At system level, Concept F1 correlates strongly with judge-measured faithfulness (Spearman 0.91) and inversely with hallucination rate (−0.93), supporting the metric as a proxy for note quality.
Model scale is secondary: the 3B lightweight system essentially matches the second-best heavyweight submission, and the gain from scaling is smaller than the gain from reinforcement learning.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The safety claim rests on a single LLM judge with no human-agreement study; a stricter human review of the supposedly low hallucination rates would tell whether the RL recipe is genuinely safe for clinical use.
The transcript-to-speech transfer result suggests a cheap general recipe for long-form audio summarization: do the expensive fine-tuning on text, then plug in a frozen speech encoder at inference; this may generalize to other domains beyond medicine.
A natural extension is to replace the F1 reward with a faithfulness-aware reward that uses the judge signal itself, which could push hallucination rates even lower while keeping concept recall.
The strong system-level metric-judge correlation does not guarantee per-note agreement; a per-dialogue correlation study on the realistic subset would test whether the metric is safe for deployment, not just for leaderboard ranking.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Desk Editor's Note

Winning BeTraC systems with a practical RL recipe, but the core safety claim leans on an external LLM judge nobody has validated.

read the letter

Worth a read if you work on medical summarization or speech LLMs. The paper does three things well. It shows a concrete recipe — zero-shot screening for long-audio robustness, LoRA SFT, then DAPO RL with the official metric as reward — that took Voxtral models to first place in both BeTraC tracks, more than doubling baseline Concept F1. It reports a clean validation-to-test transfer and a useful observation: fine-tuning on transcripts transfers to audio almost as well as audio fine-tuning, and may help under domain shift. The authors are transparent about using the validation split for model and reward selection and about the realistic subset being only three dialogues.

The soft spots are real but not disqualifying. The biggest is the hallucination-rate evidence. The low rates come from the organizers' LLM-as-judge pipeline, not from the authors' own evaluation, and there is no human agreement study, no error analysis, and no confidence intervals. The authors themselves note that the metric and the judge both derive from LLMs, which is honest but doesn't fix the concern. The conclusion that RL added C-F1 'without inflating ... hallucinations' is not directly supported: Table V shows only final systems, and no SFT-only hallucination rates are reported, so that comparison is missing. You can accept the weaker claim that RL against Concept F1 need not produce obviously high hallucination rates; you cannot yet accept the stronger causal claim.

Also minor: all comparisons are point estimates without repeated runs or significance tests. The 'no generalization gap' claim is consistent with test and validation coming from the same synthetic pipeline; the mock and realistic subsets are the real test, and the realistic subset is anecdotal.

On balance, this is a solid systems paper for a challenge venue. The central ranking claims are externally evaluated and credible. The safety conclusion needs more support, but that's a fixable weakness, not a fatal one. I'd send it to review with a request for human validation or error analysis of the judge, SFT-only hallucination numbers, and ideally code or checkpoints.

Referee Report

3 major / 5 minor

Summary. The paper describes TalTech's submissions to the Beyond Transcription Challenge (BeTraC), where systems must generate SOAP notes directly from long doctor-patient audio without an intermediate transcription component. The authors screen open-weight speech LLMs for long-audio robustness, then adapt Voxtral Mini (lightweight track) and Voxtral Small (heavyweight track) using LoRA supervised fine-tuning followed by DAPO reinforcement learning whose reward is the challenge's Open Medical Concept F1 (C-F1). Their primary systems ranked first in both tracks according to the official blind evaluation, and an LLM-as-judge evaluation reports low hallucination rates for these systems. The paper further reports that transcript-based fine-tuning transfers to audio input and that the systems remain competitive on acted and, to a limited extent, realistic test subsets.

Significance. If the results hold, the paper provides a practical recipe for direct audio-to-SOAP generation with open-weight models, and it offers evidence that optimizing a concept-matching metric with reinforcement learning need not produce factually unreliable notes. The main strengths are the objective official leaderboard results (first in both tracks), the strong validation-to-test consistency (e.g., 0.571 vs. 0.563 C-F1 for the heavyweight system), and the cross-modal transfer finding. The central safety claim, however, rests on an LLM-as-judge evaluation that is not validated against human judgments, so the factual-reliability conclusion is currently weaker than the abstract suggests. The official ranking itself is not circular, since it is an external blind evaluation against gold references, but the absence of statistical uncertainty and the tiny realistic subset limit some of the secondary claims.

major comments (3)

[§V-C, Table V] The safety-relevant claim—that RL against Concept F1 'need not compromise factual reliability'—is supported only by the Gemma 4 31B judge's hallucination and contradiction rates. No human agreement study, no judge error analysis, no confidence intervals, and no prompt/decoding details are provided for this judge. If the judge is systematically lenient toward plausible-sounding but ungrounded statements, the reported 0.08% and 0.25% rates are not evidence of factual reliability. The system-level Spearman correlations (ρ=0.91, -0.93) do not mitigate this, because both signals come from LLMs and may share blind spots. This is the single most load-bearing gap and needs to be addressed directly.
[§VI, Table V] The conclusion states that RL added 0.04–0.05 C-F1 'without inflating note length or hallucinations.' Table V reports only the final RL-trained systems; no SFT-only hallucination/contradiction rates or note lengths are shown. Table III reports validation C-F1/R-2/R-3 but not hallucination rates or lengths. Therefore the data support at most the weaker claim that the final RL-trained systems have low judged hallucination rates, not that RL did not increase hallucination relative to SFT. Please provide SFT-only judge scores and lengths, or phrase the conclusion more narrowly.
[§V-B, Table IV] The robustness-on-real-recordings claim in the abstract and §V-B relies on the Real. column of Table IV, which contains only three dialogues. The text acknowledges this is 'anecdotal,' but the abstract still asserts that the recipe 'appears to improve robustness on out-of-domain real recordings.' With n=3 and no error bars or case-level analysis, that claim is unsupported beyond anecdote. The mock subset (272 dialogues) provides somewhat stronger but still acted evidence. Please either remove the realistic-subset claim from the abstract or add appropriate uncertainty quantification and explicit hedging.

minor comments (5)

[§IV-A, Table IIa] The LoRA target modules are listed as 'all linear'; please specify the exact modules (e.g., attention and MLP projections) for reproducibility.
[Table I] The ASR/WER evaluation protocol is not described: which subset of the validation split, what normalization, and what alignment/tool were used? This would help interpret the WER values used for model screening.
[§IV-B, Table IIb] The KL β is listed as 0. Please clarify whether DAPO was run with no KL penalty and whether any divergence or reward-hacking behavior was observed during RL training.
[§IV-C] For the facts-and-SOAP contrastive model, it is unclear whether the fact table is emitted at inference time as part of the final output or suppressed. If it is emitted, how is it reconciled with the requirement to produce plain-text SOAP notes? This affects interpretation of the 306-word mean length in Table IV.
[Throughout] There are typographical and formatting issues, including broken 'V oxtral' spacing and 'aboverecall' in §V-C. The paper would benefit from a careful proofreading pass.

Circularity Check

0 steps flagged

No circularity: blind external ranking and separate LLM judge; only disclosed reward/validation overlap.

full rationale

The paper's derivation chain is not circular. The primary claims are empirical competition outcomes: the official ranking is computed by BeTraC organizers on a blind test set of 875 audio examples using Open Medical Concept F1 (Section II, Table IV), and the hallucination/faithfulness numbers in Table V come from the organizers' LLM-as-judge pipeline that verifies atomic claims against gold transcripts (Section V-C), not from any metric optimized by the system. The only self-referential element is that DAPO RL uses Concept F1 as its reward (Section IV-B) and the validation split was used for model/checkpoint/reward selection (Section V-A); this makes validation C-F1 gains partly expected, but the paper does not rename a fitted quantity as a prediction—the test-set result is external and blind. All load-bearing citations ([2] Voxtral, [3] BeTraC corpus/evaluation, [9] DAPO) are external works, not the present authors' prior results invoked to force a conclusion. The paper even flags the shared-LLM limitation of its metric-validation correlation: 'though both signals ultimately derive from LLMs and per-note agreement may be weaker' (Section V-C). The absence of human agreement data for the Gemma judge and the lack of SFT-only hallucination rates for the 'without inflating' claim are evidence gaps about reliability, not circular reductions; they affect confidence in the safety conclusion, but do not make any derivation equivalent to its inputs. Therefore no circular step.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central claims rest on the evaluation stack (metric, LLM judge, synthetic corpus) rather than on mathematical derivation. Training hyperparameters are hand-selected free parameters; the reward function is the official metric. No new entities, forces, or conserved quantities are introduced.

free parameters (4)

LoRA rank / alpha / dropout = 16 / 16 / 0.05
Hand-chosen for SFT; no sweep reported. Controls adaptation capacity and regularization.
Learning rate = 1e-5
Hand-selected; standard for LoRA SFT; affects both SFT and RL convergence.
Effective batch size / epochs / weight decay = 16 / 1.0 / 0.1
Hand-selected training schedule; no ablations are shown.
RL rollouts per sample / steps per generation / KL beta = 8 / 4 / 0
DAPO hyperparameters chosen without reported tuning; KL beta=0 means no explicit KL penalty, relying on clipping.

axioms (4)

domain assumption The Open Medical Concept F1 metric computed with MeSH keyword matching and scispaCy NER against Kimi-K2-generated reference notes is a valid measure of clinical note quality.
Used as both reward and ranking metric. The reported correlation with LLM-judge faithfulness (Spearman ρ=0.91) is computed with LLM-derived signals and does not validate the metric against human clinical judgment.
domain assumption The LLM-as-judge pipeline using Gemma 4 31B to decompose notes into atomic claims and verify against gold transcripts reliably measures hallucination and faithfulness.
Central to the claim that RL training does not increase hallucination. No human agreement or error analysis of the judge is provided.
domain assumption SynthDoPaCo's synthetic conversations are representative enough of real doctor-patient dialogue to support cross-domain robustness conclusions.
Training and EE test come from the same synthetic pipeline; the realistic subset has only 3 dialogues, so out-of-domain claims are anecdotal.
domain assumption Voxtral's internal 30-second chunking and frozen Whisper-initialized encoder preserve all clinically relevant content in long audio.
Model selection is based on Voxtral's long-audio robustness; if chunking drops context, summarization fidelity degrades. The paper does not analyze per-chunk errors.

pith-pipeline@v1.3.0-alltime-deepseek · 6205 in / 13580 out tokens · 124561 ms · 2026-08-01T18:37:47.403386+00:00 · methodology

0 comments

read the original abstract

This paper describes TalTech's submissions to the Beyond Transcription Challenge (BeTraC), which requires generating SOAP notes directly from long doctor-patient conversation recordings, without intermediate transcription. After screening open-weight speech LLMs for long-audio robustness, we adapted Voxtral Mini (lightweight track) and Voxtral Small (heavyweight track) with LoRA supervised fine-tuning followed by DAPO reinforcement learning that uses the challenge metric, Open Medical Concept F1, as its reward. Our systems ranked first in both tracks, and an independent LLM-as-a-judge evaluation showed the lowest hallucination rate among all submissions, indicating that reinforcement learning against a concept-matching metric need not compromise factual reliability. We also find that fine-tuning on text transcripts transfers well to speech input and appears to improve robustness on out-of-domain real recordings.

discussion (0)

Reference graph

Works this paper leans on

12 extracted references · 3 linked inside Pith

[1]

BeTraC: Beyond Transcription Challenge,

“BeTraC: Beyond Transcription Challenge,” https://betrac.github.io, 2026

2026
[2]

V oxtral,

Alexander H. Liu, Andy Ehrenberg, Andy Lo, Cl ´ement Denoix, Corentin Barreau, Guillaume Lample, Jean-Malo Delignon, Khyathi Raghavi Chandu, Patrick von Platen, Pavankumar Reddy Muddireddy, Sanchit Gandhi, Soham Ghosh, Srijan Mishra, Thomas Foubert, Abhinav Rastogi, Adam Yang, Albert Q. Jiang, Alexandre Sablayrolles, Am ´elie H ´eliou, Am´elie Martin, Anm...

Pith/arXiv arXiv 2025
[3]

Generating synthetic doctor-patient conversations for long-form audio summarization,

Yanis Labrak, David Gr ¨unert, S´everin Baroudi, Jiyun Chun, Pawel Cyrta, Sergio Burdisso, Ahmed Hassoon, David Liu, Adam Rothschild, Reed Van Deusen, Petr Motlicek, Andrew Perrault, Ricard Marxer, and Thomas Schaaf, “Generating synthetic doctor-patient conversations for long-form audio summarization,” inProc. Interspeech, 2026

2026
[4]

BeTraC 2026 — evaluation metrics,

“BeTraC 2026 — evaluation metrics,” https://github.com/betrac/ betrac-metrics, 2026

2026
[5]

Qwen2.5-Omni technical report,

Jin Xu, Zhifang Guo, Jinzheng He, Hangrui Hu, Ting He, Shuai Bai, Keqin Chen, Jialin Wang, Yang Fan, Kai Dang, Bin Zhang, Xiong Wang, Yunfei Chu, and Junyang Lin, “Qwen2.5-Omni technical report,”arXiv preprint arXiv:2503.20215, 2025

Pith/arXiv arXiv 2025
[6]

MOSS-Audio technical report,

Chen Yang, Chufan Yu, Hanfu Chen, Jie Zhu, Jingqi Chen, Ke Chen, Wenxuan Wang, Yang Wang, Yaozhou Jiang, Yi Jiang, Zhengyuan Lin, Ziqi Chen, Zhaoye Fei, Chenghao Liu, Jun Zhan, Kang Yu, Kexin Huang, Mingshu Chen, Qinyuan Cheng, Ruixiao Li, Shimin Li, Songlin Wang, Yang Gao, Yiyang Zhang, and Xipeng Qiu, “MOSS-Audio technical report,” 2026

2026
[7]

SWIFT: A scalable lightweight infrastructure for fine-tuning,

Yuze Zhao, Jintao Huang, Jinghan Hu, Xingjun Wang, Yunlin Mao, Daoze Zhang, Zeyinzi Jiang, Zhikai Wu, Baole Ai, Ang Wang, Wenmeng Zhou, and Yingda Chen, “SWIFT: A scalable lightweight infrastructure for fine-tuning,” 2024

2024
[8]

DeepSeekMath: Pushing the limits of mathematical reasoning in open language models,

Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y . K. Li, Y . Wu, and Daya Guo, “DeepSeekMath: Pushing the limits of mathematical reasoning in open language models,”arXiv preprint arXiv:2402.03300, 2024

Pith/arXiv arXiv 2024
[9]

DAPO: An open- source llm reinforcement learning system at scale,

Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Weinan Dai, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, Haibin Lin, Zhiqi Lin, Bole Ma, Guangming Sheng, Yuxuan Tong, Chi Zhang, Mofan Zhang, Wang Zhang, Hang Zhu, Jinhua Zhu, Jiaze Chen, Jiangjie Chen, Chengyi Wang, Hongli Yu, Yuxuan Song, Xiangpeng Wei, Hao Zhou, Jingjing Liu, W...

2025
[10]

Omni-R1: Do you really need audio to fine-tune your audio llm?,

Andrew Rouditchenko, Saurabhchand Bhati, Edson Araujo, Samuel Thomas, Hilde Kuehne, Rogerio Feris, and James Glass, “Omni-R1: Do you really need audio to fine-tune your audio llm?,” inProc. ASRU, 2025

2025
[11]

Exploring fine-tuning of large audio language models for spoken language understanding under limited speech data,

Youngwon Choi, Jaeyoon Jung, Hyeonyu Kim, Huu-Kim Nguyen, and Hwayeon Kim, “Exploring fine-tuning of large audio language models for spoken language understanding under limited speech data,” inProc. ICASSP, 2026

2026
[12]

A dataset of simulated patient-physician medical interviews with a focus on respiratory cases,

Faiha Fareez, Tishya Parikh, Christopher Wavell, Saba Shahab, Meghan Chevalier, Scott Good, Isabella De Blasi, Rafik Rhouma, Christopher McMahon, Jean-Paul Lam, Thomas Lo, and Christopher W. Smith, “A dataset of simulated patient-physician medical interviews with a focus on respiratory cases,”Scientific Data, vol. 9, no. 1, pp. 313, jun 2022

2022

[1] [1]

BeTraC: Beyond Transcription Challenge,

“BeTraC: Beyond Transcription Challenge,” https://betrac.github.io, 2026

2026

[2] [2]

V oxtral,

Alexander H. Liu, Andy Ehrenberg, Andy Lo, Cl ´ement Denoix, Corentin Barreau, Guillaume Lample, Jean-Malo Delignon, Khyathi Raghavi Chandu, Patrick von Platen, Pavankumar Reddy Muddireddy, Sanchit Gandhi, Soham Ghosh, Srijan Mishra, Thomas Foubert, Abhinav Rastogi, Adam Yang, Albert Q. Jiang, Alexandre Sablayrolles, Am ´elie H ´eliou, Am´elie Martin, Anm...

Pith/arXiv arXiv 2025

[3] [3]

Generating synthetic doctor-patient conversations for long-form audio summarization,

Yanis Labrak, David Gr ¨unert, S´everin Baroudi, Jiyun Chun, Pawel Cyrta, Sergio Burdisso, Ahmed Hassoon, David Liu, Adam Rothschild, Reed Van Deusen, Petr Motlicek, Andrew Perrault, Ricard Marxer, and Thomas Schaaf, “Generating synthetic doctor-patient conversations for long-form audio summarization,” inProc. Interspeech, 2026

2026

[4] [4]

BeTraC 2026 — evaluation metrics,

“BeTraC 2026 — evaluation metrics,” https://github.com/betrac/ betrac-metrics, 2026

2026

[5] [5]

Qwen2.5-Omni technical report,

Jin Xu, Zhifang Guo, Jinzheng He, Hangrui Hu, Ting He, Shuai Bai, Keqin Chen, Jialin Wang, Yang Fan, Kai Dang, Bin Zhang, Xiong Wang, Yunfei Chu, and Junyang Lin, “Qwen2.5-Omni technical report,”arXiv preprint arXiv:2503.20215, 2025

Pith/arXiv arXiv 2025

[6] [6]

MOSS-Audio technical report,

Chen Yang, Chufan Yu, Hanfu Chen, Jie Zhu, Jingqi Chen, Ke Chen, Wenxuan Wang, Yang Wang, Yaozhou Jiang, Yi Jiang, Zhengyuan Lin, Ziqi Chen, Zhaoye Fei, Chenghao Liu, Jun Zhan, Kang Yu, Kexin Huang, Mingshu Chen, Qinyuan Cheng, Ruixiao Li, Shimin Li, Songlin Wang, Yang Gao, Yiyang Zhang, and Xipeng Qiu, “MOSS-Audio technical report,” 2026

2026

[7] [7]

SWIFT: A scalable lightweight infrastructure for fine-tuning,

Yuze Zhao, Jintao Huang, Jinghan Hu, Xingjun Wang, Yunlin Mao, Daoze Zhang, Zeyinzi Jiang, Zhikai Wu, Baole Ai, Ang Wang, Wenmeng Zhou, and Yingda Chen, “SWIFT: A scalable lightweight infrastructure for fine-tuning,” 2024

2024

[8] [8]

DeepSeekMath: Pushing the limits of mathematical reasoning in open language models,

Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y . K. Li, Y . Wu, and Daya Guo, “DeepSeekMath: Pushing the limits of mathematical reasoning in open language models,”arXiv preprint arXiv:2402.03300, 2024

Pith/arXiv arXiv 2024

[9] [9]

DAPO: An open- source llm reinforcement learning system at scale,

Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Weinan Dai, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, Haibin Lin, Zhiqi Lin, Bole Ma, Guangming Sheng, Yuxuan Tong, Chi Zhang, Mofan Zhang, Wang Zhang, Hang Zhu, Jinhua Zhu, Jiaze Chen, Jiangjie Chen, Chengyi Wang, Hongli Yu, Yuxuan Song, Xiangpeng Wei, Hao Zhou, Jingjing Liu, W...

2025

[10] [10]

Omni-R1: Do you really need audio to fine-tune your audio llm?,

Andrew Rouditchenko, Saurabhchand Bhati, Edson Araujo, Samuel Thomas, Hilde Kuehne, Rogerio Feris, and James Glass, “Omni-R1: Do you really need audio to fine-tune your audio llm?,” inProc. ASRU, 2025

2025

[11] [11]

Exploring fine-tuning of large audio language models for spoken language understanding under limited speech data,

Youngwon Choi, Jaeyoon Jung, Hyeonyu Kim, Huu-Kim Nguyen, and Hwayeon Kim, “Exploring fine-tuning of large audio language models for spoken language understanding under limited speech data,” inProc. ICASSP, 2026

2026

[12] [12]

A dataset of simulated patient-physician medical interviews with a focus on respiratory cases,

Faiha Fareez, Tishya Parikh, Christopher Wavell, Saba Shahab, Meghan Chevalier, Scott Good, Isabella De Blasi, Rafik Rhouma, Christopher McMahon, Jean-Paul Lam, Thomas Lo, and Christopher W. Smith, “A dataset of simulated patient-physician medical interviews with a focus on respiratory cases,”Scientific Data, vol. 9, no. 1, pp. 313, jun 2022

2022