Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

Med-U1: Incentivizing Unified Medical Reasoning in LLMs via Large-scale Reinforcement Learning

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a single pure reinforcement-learning recipe, with task-specific rule-based rewards and a length penalty, can make a 7B medical QA model outperform much larger open-source and proprietary models across…

desk verdict A genuinely useful combination of pure RL, mixed rewards, and length control for medical QA, with solid empirical gains but some reporting sloppiness and an unvalidated open-text reward proxy. read the letter →

arxiv 2506.12307 v2 pith:VSBWRL6D submitted 2025-06-14 cs.CL cs.AI

Xiaotian Zhang , Yuan Wang , Zhaopeng Feng , Ruizhe Chen , Zhijie Zhou , Yan Zhang , Hongxia Xu , Jian Wu
show 1 more author
Zuozhu Liu
This is my paper
classification cs.CLcs.AI
keywords medicalquestionansweringreinforcementlearningrule-basedrewardsGRPOreasoninglengthcontrollargelanguagemodelsbenchmarksout-of-distributiongeneralization
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

Med-U1 asks whether a single training recipe—pure reinforcement learning with rule-based rewards and no extra supervised reasoning traces—can make one language model handle the full range of medical question answering. The authors answer yes: starting from Qwen2.5-3B/7B-Instruct, they train with GRPO on a mixed reward that checks output format, verifies the answer in a task-specific way, and optionally penalizes deviation from a requested reasoning length. On five medical benchmarks the 7B model surpasses much larger open-source and proprietary baselines, and on an out-of-distribution health benchmark it beats a supervised fine-tuning baseline by up to nine points. If true, this means a unified, verifiable RL objective can substitute for task-specific supervised data and expensive verifiers in medicine.

What carries the argument

The load-bearing object is the mixed task-specific reward function, made of three terms: a format reward R_format that scores whether the output contains <think></think> and <answer></answer> tags; a correctness reward R_correct whose verification rule depends on the task type, using literal option matching for multiple choice, range or string matching for numeric and short answers, and for open text a thresholded average of Rouge-L (longest-common-subsequence overlap) and Exact Match Score; and an optional length reward R_length that is maximized when the total thinking length matches a user-specified target. These rewards are combined and optimized with GRPO. The open-text thresholded proxy is what lets the RL loop treat free-form medical answers as verifiable, while the length term is what makes reasoning budgets controllable.

What would settle it

Retrain or re-evaluate Med-U1-7B on open-ended medical questions whose reference answers have several synonymous phrasings with low lexical overlap; if expert clinicians score the outputs as correct while the Rouge-L/EMS reward scores them below threshold, the reward proxy, not medical reasoning, is the binding constraint.

Watch

Extended reading notes

Core claim

The paper's central claim is that pure large-scale reinforcement learning, guided by a mixed rule-based reward, is enough to produce a unified medical QA system from a 7B instruction-tuned model. The authors report that Med-U1-7B reaches 57.55% accuracy on MedCalc-Bench and 17.74% on MedXpertQA, and that on the open-ended medical-o1-reasoning-SFT task it achieves a 73.06 Exact Match Score, surpassing larger open-source and proprietary baselines on the aggregated comparisons. They also find that the RL-trained model beats an SFT baseline by up to nine percentage points on an out-of-distribution health benchmark. The authors interpret these results as evidence that the RL paradigm itself, not the distribution of supervised reasoning traces, drives robust and generalizable medical reasoning.

Load-bearing premise

The method assumes a free-text answer is correct exactly when its Rouge-L/Exact-Match blended score passes a fixed threshold, so medically valid answers phrased differently from the reference are treated as wrong and the model is pushed away from them.

Editorial extensions

If this is right

  • A single RL run on mixed-format medical data can replace pipelines that first distill reasoning traces with SFT and then apply RL, cutting the need for human- or model-generated reasoning annotations.
  • Open-ended medical answers can be trained with fully rule-based rewards once correctness is defined as a thresholded lexical-overlap score, so no learned reward model or LLM judge is needed.
  • Reasoning length becomes a user-tunable dial: the model can be prompted with a target length, and most medical tasks reach peak accuracy near 512 tokens, so inference cost can be cut without losing accuracy.
  • Pure RL appears safer than SFT-then-RL for calculation-heavy medical tasks, where prior SFT can interfere with numerical reasoning.
  • Out-of-distribution gains on a held-out health benchmark suggest the RL-trained model learns a general problem-solving behavior rather than memorizing the five training sets.

Reading between the lines

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

  • Inference: the open-text results are only as medically trustworthy as the Rouge-L/EMS threshold; if the threshold mis-scores synonymous phrasings, the model will learn surface mimicry, so an expert-reviewed sample of near-miss outputs would show whether the proxy is sound.
  • Inference: the bell-shaped length-accuracy curves imply that per-task length targets could be learned automatically from validation data, giving a cheap adaptive inference policy the paper does not implement.
  • Inference: the same reward decomposition—format, task-specific correctness, and length—could transfer to other domains with heterogeneous outputs, such as legal or scientific QA, where answer formats are similarly mixed.
  • Inference: because the format reward scores the use of <think>/<answer> tags, part of the EMS gain may come from enforced structure rather than better medical knowledge; ablating the format reward would isolate that effect.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. Med-U1 proposes a pure reinforcement-learning framework for unified medical question answering, trained with Qwen2.5-3B/7B-Instruct backbones via GRPO. The reward is a mixed rule-based signal covering three output formats: multiple-choice options, numeric values, and open-ended text, with an optional length-matching reward to control reasoning verbosity. The model is trained on five medical datasets and evaluated on in-distribution benchmarks plus the MMLU-Pro health subset. The paper reports consistent gains over SFT baselines and several open-source and proprietary models, along with analyses of reasoning length, reward design, and training dynamics.

Significance. If the results are reproducible and robust, Med-U1 provides a useful data point that pure RL with simple rule-based rewards can unify heterogeneous medical QA formats at 7B scale, while also offering a mechanism for length control. The work's strengths include a clear and simple reward formulation, a unified training recipe across three output types, an explicit OOD evaluation, and a stated commitment to open-sourcing the code. The claims would be substantially strengthened by multi-seed reporting, a validated open-text correctness reward, and a specification of how the length target is constructed.

major comments (5)
  1. [Section 2.1 and Section 3.1] The open-text correctness reward is a thresholded average of Rouge-L and EMS, and Section 3.1 reports thresholds of 40, 50, and 70 for Rouge-L, Mix, and EMS, respectively, without describing how these thresholds were selected or validated. The paper documents a failure case for BLEU but does not provide analogous evidence that the Rouge-L/EMS threshold correctly identifies medically correct paraphrases; the Limitations section concedes that "more fine-grained and clinically grounded reward objectives" are needed. Because the same metrics are used for evaluation in Table 3, the open-text improvements may partly reflect overfitting to lexical overlap rather than medical correctness. Please report the threshold selection procedure (train/validation split vs. test set), a sensitivity analysis, and ideally a correlation of the reward with human judgments on a held-out set.
  2. [Section 3.1 and Tables 1-2] No variance or confidence intervals are reported for any experiment, and several reported differences are small or even negative (e.g., EHRNoteQA for Med-U1-3B is 55.17 vs. 52.87 for the SFT baseline, and MMLU-Pro Virology for Med-U1-3B is 36.96 vs. 43.48 for SFT). GRPO training is stochastic, so single-run accuracy is insufficient to support the consistent-gains claim. Please report results over at least three seeds with standard deviations or confidence intervals for the main tables.
  3. [Section 3.2 and Table 2] The text states that improvements on the 3B model "can reach up to 9 percentage points" and that Med-U1-7B shows "approximately 19 percentage points" gain in select OOD subdomains, but Table 2 shows much larger differences: the largest 3B gain is about 29 points (College Medicine: 47.92 vs. 18.75) and the largest 7B gain is about 23 points (College Medicine: 64.58 vs. 41.67); some subdomains show negative differences. Please reconcile the text with the table and specify whether the 9-point and 19-point figures refer to a particular subdomain, an average, or a different metric.
  4. [Section 2.2 and Section 3.1] The length-matching reward R_length in Eq. (3) requires a target length lgold for each training question, but the paper never states how lgold is constructed (e.g., from reference answers, a heuristic, or a learned prior) nor whether the main Med-U1 models in Table 1 were trained with R_length active or with the length term omitted. This is essential for reproducing the controllable-length claim and for understanding whether the main results depend on the length reward. Please specify the construction of lgold and the training configuration for the models reported in Table 1.
  5. [Appendix B.3] The evaluation section does not specify whether the same answer-extraction protocol (parsing <answer></answer> tags via regular expressions) is applied to all baselines. If baseline models are not instructed or trained to use this format, their MCQ and open-text scores may be artificially lowered by the extraction procedure. Please clarify the exact extraction steps for each model family and, if needed, provide an alternative evaluation (e.g., using the full output or the last answer span) to ensure a fair comparison.
minor comments (6)
  1. [Section 2.1] There is a typo in the sentence describing numeric value computation: "in ths category" should be "in this category".
  2. [Section 2.1] The yellow-box text says "faluire case" and should read "failure case".
  3. [Section 3.1] The threshold values 40, 50, and 70 are presumably percentages (0.40, 0.50, 0.70) for Rouge-L, Mix, and EMS, respectively; please state the units explicitly and confirm whether the thresholds are applied to the continuous scores or to the averaged score for the Mix reward.
  4. [Figure 4] The labels "+9 pts" and "+5 pts" in Figure 4 are unclear without a legend or axis explanation; please clarify what comparison these deltas refer to and whether they correspond to the 3B or 7B model.
  5. [Table 3] The table header appears garbled in the manuscript; please format it with clear columns for the reward type and the evaluation metrics (Rouge-L and EMS) so that the comparison between Rouge-L, EMS, and Mix rewards is readable.
  6. [Appendix B.4] The learning rate is written as "e-5"; this should be "1e-5".

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; rewards are defined against external ground-truth labels and evaluation is benchmark-based.

full rationale

The paper's derivation chain is empirical rather than a formal reduction: Med-U1 is trained with GRPO on a mixed reward (format, correctness, length), where correctness rewards are computed against external ground-truth labels—exact option match for MCQs, numeric interval/string match for value tasks, and a thresholded Rouge-L/EMS mix for open text. None of these rewards is defined in terms of the model's own outputs or in terms of the performance figures being reported, so the training signal is not self-referential. The open-text thresholds (tau=40/50/70) are hyperparameters of the reward, and the fact that the same metric family also appears in evaluation is a standard proxy-alignment setup, not a fitted parameter renamed as a prediction; the paper's Limitations section explicitly flags the need for more clinically grounded rewards, which is a validity caution rather than circularity. Self-citations (Jiang et al. 2024; Liu et al. 2024) appear only in background discussion and are not load-bearing for the central RL claim. Comparisons are made against external proprietary and open-source baselines on held-out benchmarks, including the OOD MMLU-Pro health subset, so the central claim retains independent empirical content.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard RL machinery and on the reliability of benchmark rewards. The main tuning burden is the set of correctness thresholds and the undefined length target; no new physical or theoretical entities are introduced.

free parameters (5)
  • Correctness threshold tau (Rouge-L) = 40
    Chosen per reward type to decide when an open-text answer counts as correct. Section 3.1.
  • Correctness threshold tau (Mix) = 50
    Chosen per reward type to decide when an open-text answer counts as correct using the averaged Rouge-L and EMS score. Section 3.1.
  • Correctness threshold tau (EMS) = 70
    Chosen per reward type to decide when an open-text answer counts as correct using Exact Match Score. Section 3.1.
  • Length reward normalization alpha = not specified
    Appears in Eq. 3 and controls the softness of the length-matching reward; its value is never reported.
  • Rouge-L / EMS mixing weight in open-text reward = 0.5 / 0.5
    Equal weighting is adopted for the Mix reward without an ablation of the weight itself. Section 2.1.
assumptions (4)
  • standard math GRPO with PPO-style clipping and KL penalty produces a valid policy gradient update
    The optimization objective in Eq. 4 relies on standard clipped importance sampling and KL regularization.
  • domain assumption Ground truth labels in EHRNoteQA, MedXpertQA, MEDiQ, MedCalc-Bench, and medical-o1-reasoning-SFT are correct and reliable
    The reward design treats these benchmark labels as the gold standard for correctness. Section 2.1.
  • domain assumption Regex extraction of answers from <answer> tags is accurate enough to assign format and correctness rewards
    Format reward depends on regular expression extraction of the predicted answer. Section 2.1.
  • ad hoc to paper Target length lgold for each training question is available or constructible
    Section 2.2 introduces lgold without specifying how it is obtained, whether from ground-truth answer length, user specification, or a heuristic.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Med-U1: Incentivizing Unified Medical Reasoning in LLMs via Large-scale Reinforcement Learning." pith.science (2026). https://pith.science/paper/VSBWRL6D

@misc{pith2026250612307,
  author       = {Pith},
  title        = {Pith review of: Med-U1: Incentivizing Unified Medical Reasoning in LLMs via Large-scale Reinforcement Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VSBWRL6D}},
  note         = {Machine review of arXiv:2506.12307}
}
read the original abstract

Medical Question-Answering (QA) encompasses a broad spectrum of tasks, including multiple choice questions (MCQ), open-ended text generation, and complex computational reasoning. Despite this variety, a unified framework for delivering high-quality medical QA has yet to emerge. Although recent progress in reasoning-augmented large language models (LLMs) has shown promise, their ability to achieve comprehensive medical understanding is still largely unexplored. In this paper, we present Med-U1, a unified framework for robust reasoning across medical QA tasks with diverse output formats, ranging from MCQs to complex generation and computation tasks. Med-U1 employs pure large-scale reinforcement learning with mixed rule-based binary reward functions, incorporating a length penalty to manage output verbosity. With multi-objective reward optimization, Med-U1 directs LLMs to produce concise and verifiable reasoning chains. Empirical results reveal that Med-U1 significantly improves performance across multiple challenging Med-QA benchmarks, surpassing even larger specialized and proprietary models. Furthermore, Med-U1 demonstrates robust generalization to out-of-distribution (OOD) tasks. Extensive analysis presents insights into training strategies, reasoning chain length control, and reward design for medical LLMs. Our code is available here.

Figures

Figures reproduced from arXiv: 2506.12307 by the authors.

Figure 1
Figure 1. Med-U1 is a length-controllable reasoning [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overall pipeline of generating specific instructions and reward scores. Figure [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Datasets overview for training and in-distribution evaluation, with examples on reward computation. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Comparing the proposed method and SFT on [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 6
Figure 6. Figure 6: Accuracy of the Med-U1-3B output on differ [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Average response length across various tasks [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Character-level length distribution of all ground truth answers in the medical-o1-reasoning-SFT (Huatuo) [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: BLEU score as the reward signal. B.2 Description of medical-o1-reasoning SFT dataset. In our text-format task, we have carefully consid￾ered the diversity of medical scenarios and the rich￾ness of medical answer formats. To achieve this, we utilized the medical-o1-reas…
Figure 10
Figure 10. Figure 10: Training dynamics under different training strategies. The red curve denotes the pure RL paradigm, [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]
Figure 11
Figure 11. Figure 11: An overview of the five datasets used in our experiments is presented, with examples drawn from each. [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: A case study comparing Qwen-Instruct-Base, Qwen-Instruct-SFT, and our Med-U1 method on MedCalc [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Aligning Clinical Needs and AI Capabilities: A Survey on LLMs for Medical Reasoning

    cs.AI 2026-07 accept novelty 6.0 of 10

    A dual clinical-computational taxonomy for medical LLM reasoning plus a five-level 5k-sample benchmark showing specialists excel at diagnosis and general models at decision support/dialogue.

Reference graph

Works this paper leans on

34 extracted references · 6 canonical work pages · cited by 1 Pith paper

  1. [1]

    Marah Abdin, Sahaj Agarwal, Ahmed Awadallah, Vidhisha Balachandran, Harkirat Behl, Lingjiao Chen, Gustavo de Rosa, Suriya Gunasekar, Mojan Javaheripi, Neel Joshi, and 1 others. 2025. Phi-4-reasoning technical report. arXiv preprint arXiv:2504.21318

    arXiv 2025

  2. [2]

    Pranjal Aggarwal and Sean Welleck. 2025. L1: Controlling how long a reasoning model thinks with reinforcement learning. arXiv preprint arXiv:2503.04697

    arXiv 2025

  3. [3]

    Bradley Butcher, Michael O’Keefe, and James Titchener. 2025. Precise length control for large language models. Natural Language Processing Journal, page 100143

    work page 2025

  4. [4]

    Junying Chen, Zhenyang Cai, Ke Ji, Xidong Wang, Wanlong Liu, Rongsheng Wang, Jianye Hou, and Benyou Wang. 2024. Huatuogpt-o1, towards medical complex reasoning with llms. arXiv preprint arXiv:2412.18925

    arXiv 2024

  5. [5]

    Maxime Griot, Coralie Hemptinne, Jean Vanderdonckt, and Demet Yuksel. 2025. Large language models lack essential metacognition for reliable medical reasoning. Nature communications, 16(1):642

    work page 2025

  6. [6]

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, and 1 others. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948

    arXiv 2025

  7. [7]

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

    arXiv 2025

  8. [8]

    Xiaoke Huang, Juncheng Wu, Hui Liu, Xianfeng Tang, and Yuyin Zhou. 2025 b . m1: Unleash the potential of test-time scaling for medical reasoning with large language models. arXiv preprint arXiv:2504.00869

    arXiv 2025

Show all 34 references
  1. [9]

    Zhongzhen Huang, Gui Geng, Shengyi Hua, Zhen Huang, Haoyang Zou, Shaoting Zhang, Pengfei Liu, and Xiaofan Zhang. 2025 c . O1 replication journey--part 3: Inference-time scaling for medical reasoning. arXiv preprint arXiv:2501.06458

    2025 arXiv

  2. [10]

    Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, and 1 others. 2024. Openai o1 system card. arXiv preprint arXiv:2412.16720

    2024 arXiv

  3. [11]

    Songtao Jiang, Tuo Zheng, Yan Zhang, Yeying Jin, Li Yuan, and Zuozhu Liu. 2024. https://arxiv.org/abs/2404.10237 Med-moe: Mixture of domain-specific experts for lightweight medical vision-language models . Preprint, arXiv:2404.10237

    2024 arXiv

  4. [12]

    Nikhil Khandekar, Qiao Jin, Guangzhi Xiong, Soren Dunn, Serina Applebaum, Zain Anwar, Maame Sarfo-Gyamfi, Conrad Safranek, Abid Anwar, Andrew Zhang, and 1 others. 2024. Medcalc-bench: Evaluating large language models for medical calculations. Advances in Neural Information Pro...

    2024

  5. [13]

    Sunjun Kweon, Jiyoun Kim, Heeyoung Kwak, Dongchul Cha, Hangyul Yoon, Kwang Kim, Jeewon Yang, Seunghyun Won, and Edward Choi. 2024. Ehrnoteqa: An llm benchmark for real-world clinical practice using discharge summaries. Advances in Neural Information Processing Systems, 37:1245...

    2024

  6. [14]

    Stella Li, Vidhisha Balachandran, Shangbin Feng, Jonathan Ilgen, Emma Pierson, Pang Wei W Koh, and Yulia Tsvetkov. 2024. Mediq: Question-asking llms and a benchmark for reliable interactive clinical reasoning. Advances in Neural Information Processing Systems, 37:28858--28888

    2024

  7. [15]

    Valentin Li \'e vin, Christoffer Egeberg Hother, Andreas Geert Motzfeldt, and Ole Winther. 2024. Can large language models reason about medical questions? Patterns, 5(3)

    2024

  8. [16]

    Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74--81

    2004

  9. [17]

    Jiaxiang Liu, Yuan Wang, Jiawei Du, Joey Tianyi Zhou, and Zuozhu Liu. 2024. Medcot: Medical chain of thought via hierarchical expert. arXiv preprint arXiv:2412.13736

    2024 arXiv

  10. [18]

    Niklas Muennighoff, Zitong Yang, Weijia Shi, Xiang Lisa Li, Li Fei-Fei, Hannaneh Hajishirzi, Luke Zettlemoyer, Percy Liang, Emmanuel Cand \`e s, and Tatsunori Hashimoto. 2025. s1: Simple test-time scaling. arXiv preprint arXiv:2501.19393

    2025 arXiv

  11. [19]

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311--318

    2002

  12. [20]

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347

    2017 arXiv

  13. [21]

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, and 1 others. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300

    2024 arXiv

  14. [22]

    Prasann Singhal, Tanya Goyal, Jiacheng Xu, and Greg Durrett. 2023. A long way to go: Investigating length correlations in rlhf. arXiv preprint arXiv:2310.03716

    2023 arXiv

  15. [23]

    Kimi Team, Angang Du, Bofei Gao, Bowei Xing, Changjiu Jiang, Cheng Chen, Cheng Li, Chenjun Xiao, Chenzhuang Du, Chonghua Liao, and 1 others. 2025. Kimi k1. 5: Scaling reinforcement learning with llms. arXiv preprint arXiv:2501.12599

    2025 arXiv

  16. [24]

    Qwen Team. 2025. Qwq-32b: Embracing the power of reinforcement learning, march 2025. URL https://qwenlm. github. io/blog/qwq-32b, page 9

    2025

  17. [25]

    Yubo Wang, Xueguang Ma, Ge Zhang, Yuansheng Ni, Abhranil Chandra, Shiguang Guo, Weiming Ren, Aaran Arulraj, Xuan He, Ziyan Jiang, and 1 others. 2024. Mmlu-pro: A more robust and challenging multi-task language understanding benchmark. In The Thirty-eight Conference on Neural I...

    2024

  18. [26]

    An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, and 1 others. 2024. Qwen2. 5 technical report. arXiv preprint arXiv:2412.15115

    2024 arXiv

  19. [27]

    Hongzhou Yu, Tianhao Cheng, Ying Cheng, and Rui Feng. 2025. Finemedlm-o1: Enhancing the medical reasoning ability of llm from supervised fine-tuning to test-time training. arXiv preprint arXiv:2501.09213

    2025 arXiv

  20. [28]

    Weizhe Yuan, Ilia Kulikov, Ping Yu, Kyunghyun Cho, Sainbayar Sukhbaatar, Jason Weston, and Jing Xu. 2024. Following length constraints in instructions. arXiv preprint arXiv:2406.17744

    2024 arXiv

  21. [29]

    Kaiyan Zhang, Sihang Zeng, Ermo Hua, Ning Ding, Zhang-Ren Chen, Zhiyuan Ma, Haoxin Li, Ganqu Cui, Biqing Qi, Xuekai Zhu, and 1 others. 2024. Ultramedical: Building specialized generalists in biomedicine. Advances in Neural Information Processing Systems, 37:26045--26081

    2024

  22. [30]

    Sheng Zhang, Qianchu Liu, Guanghui Qin, Tristan Naumann, and Hoifung Poon. 2025. Med-rlvr: Emerging medical reasoning from a 3b base model via reinforcement learning. arXiv preprint arXiv:2502.19655

    2025 arXiv

  23. [31]

    Yaowei Zheng, Richong Zhang, Junhao Zhang, Yanhan Ye, Zheyan Luo, Zhangchi Feng, and Yongqiang Ma. 2024. Llamafactory: Unified efficient fine-tuning of 100+ language models. arXiv preprint arXiv:2403.13372

    2024 arXiv

  24. [32]

    Yuxin Zuo, Shang Qu, Yifei Li, Zhangren Chen, Xuekai Zhu, Ermo Hua, Kaiyan Zhang, Ning Ding, and Bowen Zhou. 2025. Medxpertqa: Benchmarking expert-level medical reasoning and understanding. arXiv preprint arXiv:2501.18362

    2025 arXiv

  25. [33]

    online" 'onlinestring :=

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint eprinttype howpublished institution journal key month note number organization pages publisher school series title type volume year doi pubmed url lastchecked label extra.label sort.label short.list...

  26. [34]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.