Pith. sign in

REVIEW 3 major objections 5 minor 6 cited by

SARI: Structured Audio Reasoning via Curriculum-Guided Reinforcement Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a curriculum-guided reinforcement-learning recipe, built on structured chain-of-thought supervision, raises audio question-answering accuracy by 16.35% over the base model and reaches 67.08% on the MMAU test-mini…

desk verdict Structured CoT + curriculum GRPO is a plausible recipe for audio reasoning, but the unvalidated LLM judge makes the headline SOTA numbers provisional until exact-match rescoring. read the letter →

arxiv 2504.15900 v3 pith:MPVT2QMV submitted 2025-04-22 cs.CL

Cheng Wen , Tingwei Guo , Shuaijiang Zhao , Wei Zou , Xiangang Li This is my paper · ORCID
classification cs.CL
keywords reinforcementlearningGRPOaudio-languagemodelchain-of-thoughtcurriculumstructuredreasoningMMAUbenchmarkmultiple-choiceaudioQA
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks whether reinforcement learning, which sharpens reasoning in text-only language models, can sharpen reasoning in models that listen. It answers yes: by first fine-tuning an audio-language model on chain-of-thought answers, then running group-relative policy optimization (GRPO) on a 32,000-sample multiple-choice audio corpus ordered from easy to hard, the model improves by 16.35 percentage points in average accuracy over its base and reaches 67.08% on the MMAU test-mini benchmark. The authors report that structured reasoning outperforms free-form reasoning and that supervised warm-up is necessary for stable RL training. If the benchmark numbers hold up, the recipe would give audio assistants a way to reason audibly and transparently about what they hear.

What carries the argument

The engine is the GRPO reinforcement-learning update applied to a large audio-language model after a supervised warm-up. The structured reasoning format, divided into Planning, Caption, Reasoning, and Summary sections, forces the model to write down what it hears before judging it. The curriculum orders the 30,000-sample RL pool by the base model's pass rate, from easiest to hardest, so the policy collects reliable rewards early. The reward itself is simple—correct answer plus format adherence—and the KL penalty is disabled.

What would settle it

Take the MMAU test-mini items, split them into clips that overlap the 32k training set and clips that do not, and recompute accuracy on each slice; if the non-overlapping slice shows no improvement over the base model, the reported gain is memorization, not reasoning. A second check is to train with the same recipe but evaluate on a benchmark with no domain overlap and no multiple-choice cues.

Watch

Extended reading notes

Core claim

The central claim is that explicit, structured reasoning plus curriculum-guided GRPO transfers the 'think before answering' paradigm to large audio-language models. The best model, SARI, built on Qwen2.5-Omni, scores 67.08% average accuracy on MMAU test-mini, which the paper labels state-of-the-art, and 66.01% on the cross-domain MMSU set; the Qwen2-Audio-based variant beats its base by 16.35 percentage points. The ablations attribute the gain to three ingredients: supervised fine-tuning on structured chains as a warm start, GRPO with answer-correctness reward, and an easy-to-hard curriculum. The authors also report that structured chains generalize more robustly than unstructured ones and that curriculum learning accelerates convergence.

Load-bearing premise

The claim rests on MMAU test-mini being a fair measure of generalization, but the training corpus overlaps with the benchmark's domain (AudioSet and music questions), and no deduplication or overlap analysis is shown.

Editorial extensions

If this is right

  • Supervised warm-up is a prerequisite: RL from scratch on the raw audio model produces meaningless chains and no accuracy gain, so the two-stage recipe is the transferable ingredient.
  • Structured chains beat unstructured ones on the same architecture, suggesting that forcing an explicit Caption step reduces errors that come from missing audio cues.
  • Curriculum ordering matters as much as the reward: easy-to-hard GRPO converges faster and lands at a higher final accuracy than the same training on a shuffled set.
  • The gains carry over to a larger base model (Qwen2.5-Omni), where the structured variant reaches 67.08% on MMAU test-mini and 66.01% on MMSU, indicating the recipe scales with base capability.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A likely next test is whether the same recipe works for open-ended audio reasoning (no multiple-choice options) and with a reward that grades reasoning quality rather than only answer correctness.
  • The structured format's Caption step may be the key transferable mechanism: writing down what is heard before reasoning could be adapted to other modalities, such as video or sensor streams, where grounding is the bottleneck.
  • Because the dataset is built entirely from open-source captions and LLM-generated questions, the recipe is reproducible at low cost; one testable extension is whether larger or more capable caption generators further raise the ceiling.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper extends GRPO-based reinforcement learning to large audio-language models (LALMs). The authors construct a 32k-sample multiple-choice audio QA corpus from AudioSet, MusicBench, MELD, and AVQA, generate structured and unstructured chain-of-thought annotations with Qwen2.5-72B, and train via SFT followed by curriculum-guided GRPO. They report that the resulting SARI model, built on Qwen2.5-Omni, achieves 67.08% average accuracy on the MMAU test-mini benchmark, which they describe as state-of-the-art, and a 16.35-point improvement over the Qwen2-Audio-7B-Instruct base model. Ablations on Qwen2-Audio-7B-Instruct compare implicit vs. explicit reasoning, structured vs. unstructured CoT, and with vs. without curriculum learning, concluding that SFT warm-up is necessary, structured CoT generalizes better, and easy-to-hard curricula improve final performance.

Significance. If the reported numbers withstand scrutiny, this is a useful contribution: it is one of the first systematic studies of GRPO with curriculum learning in the audio modality, and the controlled comparisons of reasoning styles under identical architectures are informative. The authors also report negative results (RL from scratch fails to induce meaningful CoT), which adds credibility. However, the central quantitative claims rest on an unvalidated LLM judge and a training/test domain overlap that is acknowledged but not analyzed. The claimed SOTA and the structured-vs-unstructured advantage are therefore not yet established to the standard required for publication.

major comments (3)
  1. [§4.2 (Evaluation Metrics)] All headline numbers in Tables 1–3 depend on the Qwen2-72B-Instruct judge that 'jointly analyzes' the question, reference answer, and model outputs, yet no validation of this judge is provided: no agreement with exact-match extraction, no human correlation, and no comparison with the official MMAU protocol. This is load-bearing because the RL reward is exact option correctness, while the reported metric is a different, soft, unvalidated judgement. The risk is that the judge rewards verbose, formatted structured-CoT outputs irrespective of the actual selected option, which would inflate the structured-vs-unstructured deltas (67.08 vs. 65.90 on Qwen2.5-Omni; 65.55 vs. 63.68 on Qwen2-Audio). Please report exact-match accuracy on the extracted final answer, agreement statistics (e.g., Cohen's kappa) between the LLM judge and exact match, or adopt the official MMAU scoring procedure.
  2. [§4.2 (Train/test overlap)] The text acknowledges that MMAU test-mini 'evaluates both in-domain performance on tasks similar to our training data (since we use some AudioSet and music questions)', but it does not provide any overlap or deduplication analysis between the training corpus (AudioSet, MusicBench, MELD, AVQA) and the test set. If test clips or question templates appear in the training data, the SOTA claim (67.08%) and the 16.35-point improvement could reflect memorization or in-domain fit rather than reasoning ability. Please provide a deduplication analysis (audio clip identity, question n-gram overlap, or source-level breakdown) and, if overlap exists, report results on the non-overlapping subset separately.
  3. [Tables 1–3 and §4.2 (No variance/significance)] The paper reports a single average over four sampled responses per model, with no per-response variance, standard errors, or significance tests. The structured-vs-unstructured differences (e.g., 67.08 vs. 65.90 on Qwen2.5-Omni; 63.58 vs. 63.65 for model-d vs. model-e) are within plausible sampling noise, yet they are used to support the central claim that structured reasoning is superior. Please provide per-response results, bootstrap confidence intervals, or a paired significance test over the four samples, and state the number of test items in each subset.
minor comments (5)
  1. [Abstract and §5] The phrase '16.35% improvement in average accuracy' is an absolute difference in percentage points (65.55 vs. 49.20), not a relative percentage improvement; please use '16.35 percentage points' or the relative figure (≈33%).
  2. [Throughout] Please correct typos: 'Howover' (§2), 'as showed' (§3.1), 'outperformes' (§5), 'LL AVA-COT' (§3.1), 'V oiceBench' (ref. [32]), and 'T able' (§5).
  3. [§5 (Generalization claim)] The claim of 'robust generalization capabilities' based on MMSU is overstated for the Qwen2-Audio-7B-Instruct variant: SARI's MMSU accuracy (35.36) is slightly below the base model (35.72). Please qualify the claim or restrict it to the Qwen2.5-Omni variant, which does improve MMSU from 61.32 to 66.01.
  4. [§4.1 and Tables 2–3] Please specify whether the Qwen2.5-Omni pipeline uses the same curriculum ordering (computed from Qwen2-Audio-7B pass rates) or a model-specific ordering, since this affects reproducibility.
  5. [§3.2] The KL divergence penalty is disabled (coefficient set to 0), which is unusual for GRPO; please discuss the rationale and any observed effect on training stability.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the central claims are empirical training results evaluated on external benchmarks, not derivations that reduce to their inputs.

full rationale

The paper's derivation chain is an empirical recipe: construct a 32k multiple-choice audio QA corpus, generate structured and unstructured reasoning annotations with Qwen2.5-72B using the ground-truth answer as a hidden conditioning signal, SFT on 2k annotated examples, then GRPO on roughly 30k examples with an exact-answer reward and an easy-to-hard curriculum. None of the reported accuracy numbers is defined as an input to the training objective. The SFT supervision is answer-conditioned reasoning, but this is standard supervised training rather than a prediction that is equivalent to its input by construction. The curriculum difficulty ranking uses the baseline model's pass rate, and the RL reward is exact option correctness; neither is a renamed version of the claimed output accuracy. The evaluation uses an unvalidated Qwen2-72B-Instruct judge for MMAU, and Section 4.2 concedes that MMAU test-mini includes in-domain AudioSet and music questions with no deduplication analysis; these are benchmark-contamination and metric-validity risks, not internal circularity. Section 7 further limits generalization claims to two Qwen architectures, which is a scope limitation rather than a circular step. No load-bearing argument reduces to a self-citation: references to LLaVA-CoT, Audio-Reasoner, and R1-AQA are external prior work, and the structured format is adopted by stated design choice, not by an invoked uniqueness theorem. Therefore the appropriate circularity score is 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The central empirical claim rests on a pipeline of generated data: model-produced captions, LLM-produced reasoning, an answer-consistency check, a difficulty filter, and human-chosen hyperparameters. None of these are independently validated, and the benchmark overlap leaves the central comparison only partially secured.

free parameters (4)
  • DataRL difficulty filter threshold = 0% pass rate across 16 attempts
    Questions that the Qwen2-Audio-7B-Instruct baseline never answers correctly in 16 trials are discarded as ambiguous or too difficult; this threshold directly determines the RL training distribution (Section 3.1).
  • SFT training schedule = 3 epochs, batch size 64, learning rate 2e-5
    Chosen without ablations; defines how strongly the model imprints the LLM-generated reasoning style before RL (Section 4.1).
  • GRPO training schedule = 1 epoch, batch size 32, learning rate 1e-6, KL coefficient 0, temperature 1.0, 4 responses per sample
    These RL hyperparameters are set by hand and the KL penalty is disabled; training stability and the reported gains depend on this configuration (Section 4.1).
  • Number of responses for pass-rate ranking = 16
    Difficulty ordering and the 0% filter both assume 16 samples is enough to estimate baseline pass rate reliably (Section 3.1).
assumptions (4)
  • domain assumption Text captions generated by Qwen2-Audio-7B-Instruct are accurate enough to support question generation and reasoning annotation.
    Stage 1 uses model-generated captions as the sole audio representation when synthesizing QA and reasoning chains; no human verification or caption accuracy audit is reported.
  • ad hoc to paper Reasoning chains generated by Qwen2.5-72B are valid as long as the extracted final answer matches the ground-truth option.
    Stage 2 verifies only final-answer consistency, not the factual correctness of intermediate reasoning steps; flawed intermediate logic can be encoded into the SFT supervision.
  • domain assumption MMAU test-mini is a valid held-out benchmark for measuring generalization despite training on overlapping data sources (AudioSet, MusicBench).
    Section 4.2 acknowledges the test set is in-domain for AudioSet and music; no deduplication or contamination analysis is provided, so SOTA numbers may not measure generalization.
  • domain assumption LLM-based grading by Qwen2-72B-Instruct is a faithful substitute for exact answer matching on MMAU test-mini.
    Section 4.2 uses an LLM judge to jointly analyze question, reference answer, and model output; no validation of judge accuracy or agreement with official evaluation is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SARI: Structured Audio Reasoning via Curriculum-Guided Reinforcement Learning." pith.science (2026). https://pith.science/paper/MPVT2QMV

@misc{pith2026250415900,
  author       = {Pith},
  title        = {Pith review of: SARI: Structured Audio Reasoning via Curriculum-Guided Reinforcement Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MPVT2QMV}},
  note         = {Machine review of arXiv:2504.15900}
}
read the original abstract

Recent work shows that reinforcement learning(RL) can markedly sharpen the reasoning ability of large language models (LLMs) by prompting them to "think before answering." Yet whether and how these gains transfer to audio-language reasoning remains largely unexplored. We extend the Group-Relative Policy Optimization (GRPO) framework from DeepSeek-R1 to a Large Audio-Language Model (LALM), and construct a 32k sample multiple-choice corpus. Using a two-stage regimen supervised fine-tuning on structured and unstructured chains-of-thought, followed by curriculum-guided GRPO, we systematically compare implicit vs. explicit, and structured vs. free form reasoning under identical architectures. Our structured audio reasoning model, SARI (Structured Audio Reasoning via Curriculum-Guided Reinforcement Learning), achieves a 16.35% improvement in average accuracy over the base model Qwen2-Audio-7B-Instruct. Furthermore, the variant built upon Qwen2.5-Omni reaches state-of-the-art performance of 67.08% on the MMAU test-mini benchmark. Ablation experiments show that on the base model we use: (i) SFT warm-up is important for stable RL training, (ii) structured chains yield more robust generalization than unstructured ones, and (iii) easy-to-hard curricula accelerate convergence and improve final performance. These findings demonstrate that explicit, structured reasoning and curriculum learning substantially enhances audio-language understanding.

Figures

Figures reproduced from arXiv: 2504.15900 by the authors.

Figure 1
Figure 1. Benchmark Performance of MMAU Test-mini & MMSU [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Data Construction 3 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Token length of different Think Pattern 3.2 Training Strategies for Reasoning Enhancement We employ a two-phase training approach for our audio-language model: first supervised fine-tuning (SFT) on the reasoning-augmented data, and then reinforcement learning (RL) fine-tuning with a re￾ward for correct format and answer. Within this framework, we explore several experimental settings to evaluate the impact of explic… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Completion length of our models derived from Qwen2-Audio-7B-Instruct in convergence [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Kalman Meets Curriculum: Efficient Dynamic Prompt Selection for Adaptive RL Finetuning

    cs.LG 2026-07 conditional novelty 6.0 of 10

    A Kalman filter with policy-update-coupled process noise tracks non-stationary prompt difficulty and selects intermediate-difficulty batches for RL finetuning without extra rollouts.

  2. MMAR: A Challenging Benchmark for Deep Reasoning in Speech, Audio, Music, and Their Mix

    cs.SD 2025-05 conditional novelty 6.0 of 10

    MMAR, a 1,000-question benchmark built from real-world videos, shows open-source audio-language models perform near chance while the best closed model reaches about 66% accuracy.

  3. Group Relative Policy Optimization for Speech Recognition

    eess.AS 2025-09 conditional novelty 5.0 of 10

    Applying GRPO with rule-based rewards to LLM-based ASR improves WER by up to 18.4% relative and reduces hallucination errors on unseen acoustic conditions.

  4. Progressive Mastery: Customized Curriculum Learning with Guided Prompting for Mathematical Reasoning

    cs.CL 2025-06 conditional novelty 4.0 of 10

    CCL orders LLM training data by the model's own measured accuracy and converts the hardest problems into hinted completion tasks, reporting higher average benchmark scores than uniform training.

  5. Reinforced MLLM: A Survey on RL-Based Reasoning in Multimodal Large Language Models

    cs.AI 2025-04 conditional novelty 4.0 of 10

    A structured survey of reinforcement-learning-based reasoning methods for multimodal large language models, including a taxonomy, reward design review, benchmark tables, and open challenges.

  6. Reinforcement Fine-Tuning Powers Reasoning Capability of Multimodal Large Language Models

    cs.CL 2025-05 conditional novelty 2.0 of 10

    A survey-style position paper claims that reinforcement fine-tuning powers reasoning in multimodal LLMs, summarizing over a hundred recent works and proposing five future research directions.

Reference graph

Works this paper leans on

41 extracted references · 12 canonical work pages · cited by 6 Pith papers

  1. [1]

    Chain-of-thought prompting elicits reasoning in large language models

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824–24837, 2022

    2022

  2. [2]

    Towards reasoning era: A survey of long chain-of-thought for reasoning large language models.arXiv preprint arXiv:2503.09567, 2025

    Qiguang Chen, Libo Qin, Jinhao Liu, Dengyun Peng, Jiannan Guan, Peng Wang, Mengkang Hu, Yuhang Zhou, Te Gao, and Wangxiang Che. Towards reasoning era: A survey of long chain-of-thought for reasoning large language models.arXiv preprint arXiv:2503.09567, 2025

    arXiv 2025

  3. [3]

    Measuring and narrowing the compositionality gap in language models

    Ofir Press, Muru Zhang, Sewon Min, Ludwig Schmidt, Noah A Smith, and Mike Lewis. Measuring and narrowing the compositionality gap in language models. arXiv preprint arXiv:2210.03350, 2022

    arXiv 2022

  4. [4]

    Encouraging divergent thinking in large language models through multi-agent debate

    Tian Liang, Zhiwei He, Wenxiang Jiao, Xing Wang, Yan Wang, Rui Wang, Yujiu Yang, Shum- ing Shi, and Zhaopeng Tu. Encouraging divergent thinking in large language models through multi-agent debate. arXiv preprint arXiv:2305.19118, 2023

    arXiv 2023

  5. [5]

    Openai o1 system card

    Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, et al. Openai o1 system card. arXiv preprint arXiv:2412.16720, 2024

    arXiv 2024

  6. [6]

    Kimi Team, Angang Du, Bofei Gao, Bowei Xing, Changjiu Jiang, Cheng Chen, Cheng Li, Chenjun Xiao, Chenzhuang Du, Chonghua Liao, et al. Kimi k1. 5: Scaling reinforcement learning with llms. arXiv preprint arXiv:2501.12599, 2025

    arXiv 2025

  7. [7]

    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, YK Li, Y Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300, 2024

    arXiv 2024

  8. [8]

    Thinkpatterns-21k: A systematic study on the impact of thinking patterns in llms

    Pengcheng Wen, Jiaming Ji, Chi-Min Chan, Juntao Dai, Donghai Hong, Yaodong Yang, Sirui Han, and Yike Guo. Thinkpatterns-21k: A systematic study on the impact of thinking patterns in llms. arXiv preprint arXiv:2503.12918, 2025

    arXiv 2025

Show all 41 references
  1. [9]

    Audio flamingo: A novel audio language model with few-shot learning and dialogue abilities

    Zhifeng Kong, Arushi Goel, Rohan Badlani, Wei Ping, Rafael Valle, and Bryan Catanzaro. Audio flamingo: A novel audio language model with few-shot learning and dialogue abilities. arXiv preprint arXiv:2402.01831, 2024

    2024 arXiv

  2. [10]

    Salmonn: Towards generic hearing abilities for large language models.arXiv preprint arXiv:2310.13289, 2023

    Changli Tang, Wenyi Yu, Guangzhi Sun, Xianzhao Chen, Tian Tan, Wei Li, Lu Lu, Zejun Ma, and Chao Zhang. Salmonn: Towards generic hearing abilities for large language models.arXiv preprint arXiv:2310.13289, 2023

    2023 arXiv

  3. [11]

    Qwen2-audio technical report

    Yunfei Chu, Jin Xu, Qian Yang, Haojie Wei, Xipin Wei, Zhifang Guo, Yichong Leng, Yuan- jun Lv, Jinzheng He, Junyang Lin, et al. Qwen2-audio technical report. arXiv preprint arXiv:2407.10759, 2024

    2024 arXiv

  4. [12]

    Audio-reasoner: Improving reasoning capability in large audio language models

    Zhifei Xie, Mingbao Lin, Zihang Liu, Pengcheng Wu, Shuicheng Yan, and Chunyan Miao. Audio-reasoner: Improving reasoning capability in large audio language models. arXiv preprint arXiv:2503.02318, 2025

    2025

  5. [13]

    Reinforce- ment learning outperforms supervised fine-tuning: A case study on audio question answering

    Gang Li, Jizhong Liu, Heinrich Dinkel, Yadong Niu, Junbo Zhang, and Jian Luan. Reinforce- ment learning outperforms supervised fine-tuning: A case study on audio question answering. arXiv preprint arXiv:2503.11197, 2025

    2025 arXiv

  6. [14]

    Curriculum learning

    Yoshua Bengio, Jérôme Louradour, Ronan Collobert, and Jason Weston. Curriculum learning. In Proceedings of the 26th annual international conference on machine learning, pages 41–48, 2009

    2009

  7. [15]

    Mmau: A massive multi-task audio understanding and reasoning benchmark

    S Sakshi, Utkarsh Tyagi, Sonal Kumar, Ashish Seth, Ramaneswaran Selvakumar, Oriol Nieto, Ramani Duraiswami, Sreyan Ghosh, and Dinesh Manocha. Mmau: A massive multi-task audio understanding and reasoning benchmark. arXiv preprint arXiv:2410.19168, 2024

    arXiv

  8. [16]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948, 2025

    2025 arXiv

  9. [17]

    Search- r1: Training llms to reason and leverage search engines with reinforcement learning

    Bowen Jin, Hansi Zeng, Zhenrui Yue, Dong Wang, Hamed Zamani, and Jiawei Han. Search- r1: Training llms to reason and leverage search engines with reinforcement learning. arXiv preprint arXiv:2503.09516, 2025. 11

    2025 arXiv

  10. [18]

    Lmm-r1: Empowering 3b lmms with strong reasoning abilities through two-stage rule-based rl

    Yingzhe Peng, Gongrui Zhang, Miaosen Zhang, Zhiyuan You, Jie Liu, Qipeng Zhu, Kai Yang, Xingzhong Xu, Xin Geng, and Xu Yang. Lmm-r1: Empowering 3b lmms with strong reasoning abilities through two-stage rule-based rl. arXiv preprint arXiv:2503.07536, 2025

    2025 arXiv

  11. [19]

    Open r1: A fully open reproduction of deepseek-r1, January 2025

    Hugging Face. Open r1: A fully open reproduction of deepseek-r1, January 2025

    2025

  12. [20]

    Logic-rl: Unleashing llm reasoning with rule-based reinforcement learning

    Tian Xie, Zitian Gao, Qingnan Ren, Haoming Luo, Yuqian Hong, Bryan Dai, Joey Zhou, Kai Qiu, Zhirong Wu, and Chong Luo. Logic-rl: Unleashing llm reasoning with rule-based reinforcement learning. arXiv preprint arXiv:2502.14768, 2025

    2025 arXiv

  13. [21]

    Dapo: An open-source llm reinforcement learning system at scale

    Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, et al. Dapo: An open-source llm reinforcement learning system at scale. arXiv preprint arXiv:2503.14476, 2025

    2025 arXiv

  14. [22]

    R1-onevision: Advancing generalized multi- modal reasoning through cross-modal formalization

    Yi Yang, Xiaoxuan He, Hongkun Pan, Xiyan Jiang, Yan Deng, Xingtao Yang, Haoyu Lu, Dacheng Yin, Fengyun Rao, Minfeng Zhu, et al. R1-onevision: Advancing generalized multi- modal reasoning through cross-modal formalization. arXiv preprint arXiv:2503.10615, 2025

    2025 arXiv

  15. [23]

    Video-r1: Reinforcing video reasoning in mllms

    Kaituo Feng, Kaixiong Gong, Bohao Li, Zonghao Guo, Yibing Wang, Tianshuo Peng, Benyou Wang, and Xiangyu Yue. Video-r1: Reinforcing video reasoning in mllms. arXiv preprint arXiv:2503.21776, 2025

    2025 arXiv

  16. [24]

    Vision-r1: Incentivizing reasoning capability in multimodal large language mod- els

    Wenxuan Huang, Bohan Jia, Zijie Zhai, Shaosheng Cao, Zheyu Ye, Fei Zhao, Yao Hu, and Shaohui Lin. Vision-r1: Incentivizing reasoning capability in multimodal large language mod- els. arXiv preprint arXiv:2503.06749, 2025

    2025 arXiv

  17. [25]

    Audio-cot: Exploring chain-of-thought reasoning in large audio language model

    Ziyang Ma, Zhuo Chen, Yuping Wang, Eng Siong Chng, and Xie Chen. Audio-cot: Exploring chain-of-thought reasoning in large audio language model. arXiv preprint arXiv:2501.07246, 2025

    2025 arXiv

  18. [26]

    Audio set: An ontology and human-labeled dataset for audio events

    Jort F Gemmeke, Daniel PW Ellis, Dylan Freedman, Aren Jansen, Wade Lawrence, R Chan- ning Moore, Manoj Plakal, and Marvin Ritter. Audio set: An ontology and human-labeled dataset for audio events. In 2017 IEEE international conference on acoustics, speech and signal processing...

    2017

  19. [27]

    Mustango: Toward controllable text-to-music generation

    Jan Melechovsky, Zixun Guo, Deepanway Ghosal, Navonil Majumder, Dorien Herremans, and Soujanya Poria. Mustango: Toward controllable text-to-music generation. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: H...

    2024

  20. [28]

    Meld: A multimodal multi-party dataset for emotion recognition in conversa- tions

    Soujanya Poria, Devamanyu Hazarika, Navonil Majumder, Gautam Naik, Erik Cambria, and Rada Mihalcea. Meld: A multimodal multi-party dataset for emotion recognition in conversa- tions. arXiv preprint arXiv:1810.02508, 2018

    2018 arXiv

  21. [29]

    Avqa: A dataset for audio-visual question answering on videos

    Pinci Yang, Xin Wang, Xuguang Duan, Hong Chen, Runze Hou, Cong Jin, and Wenwu Zhu. Avqa: A dataset for audio-visual question answering on videos. In Proceedings of the 30th ACM international conference on multimedia, pages 3480–3491, 2022

    2022

  22. [30]

    An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, et al. Qwen2. 5 technical report. arXiv preprint arXiv:2412.15115, 2024

    2024 arXiv

  23. [31]

    Llava-cot: Let vision language models reason step-by-step, 2025

    Guowei Xu, Peng Jin, Hao Li, Yibing Song, Lichao Sun, and Li Yuan. Llava-cot: Let vision language models reason step-by-step, 2025

    2025

  24. [32]

    V oicebench: Benchmarking llm-based voice assistants

    Yiming Chen, Xianghu Yue, Chen Zhang, Xiaoxue Gao, Robby T Tan, and Haizhou Li. V oicebench: Benchmarking llm-based voice assistants. arXiv preprint arXiv:2410.17196, 2024

    2024 arXiv

  25. [33]

    de- scription,

    Sakshi Gupta. Mmau benchmark homepage. https://sakshi113.github.io/ mmau_homepage/, 2024. 12 A PROMPTS A.1 Prompts for Reasoning Path Generation Prompt for Structured Thought Generation (Model: Qwen2.5-72B-Instruct) We are now designing a system to generate audio-based chain-o...

    2024

  26. [34]

    **Identify the Whoop**: The whoop is a loud, sharp sound that is typically associated with human vocalizations

  27. [35]

    **Compare with Human Whoops**: Human whoops can vary in intensity and frequency but are generally produced by individuals expressing strong emotions such as joy, excite- ment, or surprise

  28. [36]

    **Compare with Animal Whoops**: Birds can produce chirping or squawking sounds, but these are usually more melodic and less sharp than a human whoop

  29. [37]

    **Compare with Machine Whoops**: Machines can produce various sounds, but these 15 are usually more mechanical and lack the human quality of a whoop

  30. [38]

    </REASONING> <SUMMARY> Based on the characteristics of the whoop and the typical sounds produced by humans, the most plausible source of the whoop is a human

    **Consider Context**: The whoop is often heard in social gatherings or during celebrations, which further supports the idea that it is a human vocalization. </REASONING> <SUMMARY> Based on the characteristics of the whoop and the typical sounds produced by humans, the most pla...

  31. [39]

    - **Human V oices and Whoops**: The presence of human voices and whoops strongly 16 suggests that the sound is coming from humans

    **Identify the Sound**: - **Flowing Water**: The audio includes the sound of flowing water, indicating a natural or recreational setting near a body of water. - **Human V oices and Whoops**: The presence of human voices and whoops strongly 16 suggests that the sound is coming ...

  32. [40]

    - **Bird**: There are no bird sounds mentioned in the audio description

    **Evaluate the Options**: - **Human**: The audio clearly features human voices and whoops. - **Bird**: There are no bird sounds mentioned in the audio description. - **Dog**: No dog sounds are mentioned. - **Machine**: There are no mechanical sounds indicated in the audio description

  33. [41]

    </REASONING> <SUMMARY> Based on the analysis of the audio clip, the whoop is most likely coming from a human source

    **Conclusion Based on Analysis**: - The audio primarily consists of human voices and whoops, making ¨Human¨the most likely source of the whoop. </REASONING> <SUMMARY> Based on the analysis of the audio clip, the whoop is most likely coming from a human source. </SUMMARY> </THI...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.