Pith. sign in

REVIEW 5 major objections 5 minor 32 references

RadReason: Radiology Report Evaluation Metric with Reasons and Sub-Scores

T0 review · 5 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read RadReason trains an offline 7B model to score radiology reports on six clinical error types with written justifications, matching GPT-4-based evaluators on expert alignment.

desk verdict Plausible offline evaluator for radiology reports with sub-scores and reasons, but the ReXVal evidence is statistically thinner than the abstract claims; worth refereeing with demands for rigor. read the letter →

arxiv 2508.15464 v1 pith:PTMYVQCO submitted 2025-08-21 cs.CL cs.AI

Yingshu Li , Yunyi Liu , Lingqiao Liu , Lei Wang , Luping Zhou This is my paper
classification cs.CLcs.AI
keywords radiologyreportevaluationGRPOreinforcementlearninginterpretablemetricReXValRadCliQsub-scorepredictionclinicalNLP
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

The paper claims that an offline, 7-billion-parameter language model can grade radiology reports with the same level of expert agreement as GPT-4-based evaluators, while also explaining each grade. The proposed metric, RadReason, outputs six sub-scores matched to clinically defined error types—false prediction, omission, incorrect location, incorrect severity, and comparison-related errors—plus a short written reason for each score. Training uses Group Relative Policy Optimization, modified by two mechanisms: Sub-score Dynamic Weighting, which re-weights the six error dimensions according to the model's current per-type F1, and Majority-Guided Advantage Scaling, which strengthens learning on hard-to-grade reports and on easy reports where the model errs. On the ReXVal benchmark, RadReason reaches a Kendall's Tau of 0.730 and a Spearman correlation of 0.871 with expert ratings, above all prior offline metrics and close to GPT-4-based CheXprompt (0.750) and FineRadScore (0.737). The practical stake is that report evaluation could run locally, with reasons, instead of sending confidential patient text to commercial APIs.

What carries the argument

Group Relative Policy Optimization (GRPO), a reinforcement-learning objective that compares several completions of the same prompt and converts their rewards into relative advantages, forms the training engine. RadReason's key modifications are (1) a smooth Gaussian sub-score reward, so a prediction close to the radiologist label gets nonzero reward; (2) Sub-score Dynamic Weighting, which uses live F1 gaps to reallocate reward weight toward error types the model is currently worst at; and (3) Majority-Guided Advantage Scaling, which uses the majority vote of a completion group against ground truth to gauge prompt difficulty and scales policy gradients accordingly. The output contract is a <t

What would settle it

Take a held-out set of, say, 100 candidate-reference report pairs independently annotated by several board-certified radiologists on the same six error types, and compare RadReason's sub-scores and overall ranking with those annotations. If its Kendall's Tau on that set falls toward the older offline baselines rather than near the GPT-4 evaluators, or if the stated reason for a high-scoring error does not match the sub-score, the parity claim fails. A cheaper probe is controlled perturbation: change one finding per report (add a hallucinated effusion, delete a lobectomy history) and check that

Watch

Extended reading notes

Core claim

RadReason's central claim is that expert-level evaluation of generated radiology reports can be learned by a small open model, not bought from a closed API. The model is trained to predict six sub-scores from the RadCliQ error taxonomy and to justify each with natural language. Training runs on GRPO with three rewards: a smooth Gaussian accuracy reward that gives partial credit for near-correct sub-scores, a structured-reasoning reward, and a format reward. Two mechanisms reshape the learning signal: Sub-score Dynamic Weighting re-weights error dimensions by current F1, and Majority-Guided Advantage Scaling estimates prompt difficulty by majority vote over sampled completions, amplifying cor

Load-bearing premise

The load-bearing premise is that roughly 4,000 synthetic reports produced by GPT-4 injecting controlled errors into 1,000 real MIMIC-CXR reports, labeled with error counts, faithfully represent how radiologists assign RadCliQ sub-scores to real report errors.

Editorial extensions

If this is right

  • If the ReXVal result transfers, hospitals and research groups can evaluate generated reports locally with a 7B model, keeping patient data inside their own infrastructure instead of sending it to cloud LLM APIs.
  • The six sub-scores give failure attribution: when a report generator degrades, developers can see whether errors are new false findings, omitted findings, wrong locations, wrong severity, or comparison errors, and retrain accordingly.
  • The same GRPO-plus-dynamic-weighting recipe is argued to extend beyond chest X-rays and beyond this taxonomy, to other imaging modalities and structured clinical outputs.
  • The benchmark numbers place RadReason between the best offline regressor (RadCliQ-v1, Kendall 0.631) and the best online GPT-4 evaluators (CheXprompt 0.750, FineRadScore 0.737), narrowing the gap between closed commercial evaluation and open local evaluation.

Reading between the lines

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

  • Because the training labels come from GPT-4's injected error counts rather than from radiologists, the paper leaves open whether dynamic weighting measured on synthetic F1 would also improve a model trained on human-annotated error profiles; that is a testable next step.
  • The majority-vote difficulty signal could double as an active-learning selector: prompts where sampled completions disagree most are the ones where expert annotation would add the most information.
  • The reasoning traces are a side benefit only if radiologists actually find them coherent; the paper does not measure whether a human reader matches the stated sub-score to the written justification, which is the most direct user-study extension.
  • If the metric is used to compare report generators, total-score parity with GPT-4 does not guarantee per-type parity on rarer, higher-stakes error categories, a point the paper does not address.
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 / 5 minor

Summary. RadReason proposes an offline 7B-LM metric for radiology report evaluation that outputs six RadCliQ-style error sub-scores plus natural-language reasoning justifications. Training uses GRPO on roughly 4,000 GPT-4-synthesized report/reference pairs with prompt-injected error counts; two mechanisms, Sub-score Dynamic Weighting (SDW) and Majority-Guided Advantage Scaling (MGAS), are introduced to reweight the reward signal and modulate policy-gradient advantages. On the 200-pair ReXVal benchmark, the method reports Kendall's tau of 0.730 and Spearman correlation of 0.871, claiming to surpass all prior offline metrics and to reach parity with online GPT-4-based evaluators.

Significance. If the empirical claims held, RadReason would be a practically valuable contribution: an offline, 7B, interpretable evaluator with six sub-scores and explanations, at a plausible parity with online GPT-4 evaluators, with lower cost and fewer privacy concerns. The paper deserves credit for evaluating on the external ReXVal human-annotation benchmark, for isolating GRPO/SDW/MGAS contributions via an ablation study, and for directly targeting a deployment-relevant problem. However, the headline claims are currently supported only by point estimates on 200 pairs, and the training labels are unvalidated synthetic labels; these issues must be resolved before the significance can be fully assessed.

major comments (5)
  1. [§4.3, Table 1] The central claim that RadReason 'surpasses all prior offline metrics and achieves parity with GPT-4-based evaluations' rests on point estimates of Kendall's tau and Spearman correlation computed on 200 ReXVal pairs, with no confidence intervals or significance tests. With n=200, the approximate standard error of Kendall's tau is ~0.048 and of Spearman is ~0.071; RadReason's 0.730/0.871 therefore has wide intervals that overlap RadCliQ-v1 (0.631/0.816) and likely GREEN (0.640). Moreover, the GPT-4 rows are explicitly labeled 'not strictly comparable' in Table 1, yet the abstract and conclusion use them to claim parity. Please report bootstrap CIs over ReXVal pairs/studies, run a paired significance test against RadCliQ-v1, and either implement the GPT-4 baselines under the same protocol or soften the parity claim.
  2. [§4.1, Eqs. (3), (10)] All GRPO reward signals are computed against GPT-4-injected error counts in 3,968 synthetic reports, not against radiologist annotations. The prompt instructs GPT-4 to generate reports with 0-1, 2-3, or 4+ errors, but there is no evidence that these counts correspond to the RadCliQ sub-score values a radiologist would assign, nor that GPT-4's error profiles match real clinical report errors. Since Eq. (3) and the MGAS ground truth in Eq. (10) are trained on these labels, the high ReXVal correlation could reflect benchmark-specific statistics rather than clinical alignment. Add a human validation set (or a sample of ReXVal expert labels) demonstrating agreement between synthetic labels and expert RadCliQ annotations, or retrain with a small amount of expert-annotated data.
  3. [Eq. (8), Algorithm 1] The SDW weights are described as 'a normalized set of weights' but Eq. (8) as printed is w_j = 1 + exp(alpha*Delta_j)/sum_k exp(alpha*Delta_k), so sum_j w_j = K+1 and they are not normalized. If a normalized weighted average was intended in Eq. (9), the denominator should include the (1 + exp(...)) terms. As written, the '1+' creates a residual weighting rather than a normalized softmax weight. Please correct the equation/algorithm and clarify which version was used in the ablation, since SDW is one of the two key contributions.
  4. [Table 2] The paper advertises fine-grained sub-scores across six error types, but the per-category correlations in Table 2 are weak for several dimensions: final model 'Absence of comparison' Kendall=0.186, Spearman=0.196, and 'Omission of comparison' Kendall=0.238, Spearman=0.252; 'Incorrect location' is moderate at 0.473/0.506. The overall total correlation of 0.730/0.871 can mask poor per-aspect alignment. The authors should report per-category confidence intervals and discuss whether these low values reflect annotation sparsity in ReXVal or a genuine limitation of the sub-score predictions; this is load-bearing for the 'fine-grained, clinically grounded' claim.
  5. [§4.2] The proposed mechanisms depend on several hyperparameters that are never given: the SDW update interval M, the SDW temperature alpha, and the MGAS exponent beta. Only phi- = 0.8, phi+ = 1.2, and c = 0.5 are reported. Since the ablations in Table 2 attribute gains to SDW and MGAS, omitting these values makes the exact training procedure unreproducible and prevents sensitivity analysis. Please report all hyperparameters and, preferably, a small sensitivity study over alpha, beta, and M.
minor comments (5)
  1. [§3.2] The structured-reasoning reward only checks for the presence of six headings via regex; a model can satisfy it by emitting template text without genuine reasoning. The qualitative examples are encouraging but do not quantify reasoning faithfulness. Consider adding a human or LLM-based faithfulness check.
  2. [§4.5, Figure 3] The color highlights in the case study are not explained in the caption; please specify which color corresponds to false prediction, omission, etc.
  3. [References] The GREEN reference entry contains 'Arne Md', which appears to be a garbled author name; please check.
  4. [§5] The conclusion's final sentence is incomplete ('not only surpasses prior metrics.'); please finish the sentence.
  5. [§4.1] The training data section says 3,968 'labeled report completions' from 1,000 anchors, but does not state how many completions per anchor or how the anchors were split between training/validation; add this detail.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the metric is trained on synthetic labels and tested on external ReXVal human annotations; only a minor self-citation appears in the synthetic-data rationale.

full rationale

RadReason is a trained reward model, not a metric derived from the evaluation benchmark. The reward signals (Eq. 3–7) compare predicted sub-scores to synthetic ground-truth labels generated from 1,000 MIMIC-CXR anchors with GPT-4-injected errors; they do not use ReXVal human scores. The headline result is a correlation computed on the external ReXVal human annotations (Table 1), so the central claim is not defined in terms of the thing it predicts and is not forced by construction. The only circularity-adjacent item is a self-citation to the authors' MRScore paper (Liu et al., 2024) in Section 4.1, used to support the premise that GPT-4 aligns closely with expert radiologists. This is a minor, non-load-bearing self-citation: the synthetic-data approach is an empirical assumption, and the final ReXVal evaluation provides an independent check. No equation in the paper reduces the predicted correlations to the training labels by construction, and no fitted parameter is renamed as a prediction. Statistical concerns about the 200-pair benchmark and the 'not strictly comparable' online baselines are evidence-strength issues, not circularity.

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

No new physical or conceptual entities are introduced; the six error categories and the benchmark come from prior work (RadCliQ and ReXVal). The central claim rests on five free hyperparameters, two domain assumptions about the benchmark and taxonomy, and one ad hoc assumption that GPT-4 synthetic labels are clinically valid.

free parameters (5)
  • Gaussian reward sigma = 0.5
    Hand-chosen tolerance for the squared-distance reward in Eqs. 3 and 5; not justified or swept.
  • SDW temperature alpha = not reported
    Controls how sharply weights focus on weak dimensions in Eq. 8; value omitted from Section 4.2.
  • SDW update interval M = not reported
    Frequency of F1 recomputation in Algorithm 1; omitted from experimental settings.
  • MGAS exponent beta = not reported
    Controls sharpness of advantage scaling in Eq. 11; omitted from experimental settings.
  • MGAS scaling bounds and threshold = phi- = 0.8, phi+ = 1.2, c = 0.5
    Hand-set constants in Eq. 11 reported in Section 4.2; no sensitivity analysis.
assumptions (5)
  • standard math GRPO advantage normalization and loss are correct and applicable to this setting.
    Adopted from Guo et al. 2025 and Shao et al. 2024; not derived here.
  • domain assumption The six RadCliQ error categories are a clinically sufficient decomposition of report errors.
    Used as the fixed scoring taxonomy in Sections 1 and 3.2; inherited from Yu et al. 2023a and not validated in this paper.
  • domain assumption ReXVal expert annotations are an accurate ground truth for clinical report quality.
    The sole evaluation benchmark (Section 4.1); 200 pairs, six radiologists, no adjudication or interval estimates.
  • ad hoc to paper GPT-4 synthetic reports with injected error counts provide valid training signals for clinical sub-scores.
    Section 4.1 builds all 3,968 training examples this way; no human validation of the injected labels or comparison to real error distributions.
  • ad hoc to paper Majority voting of discrete sub-score predictions is a sound estimator of prompt difficulty.
    MGAS in Eq. 10 treats mode-vs-ground-truth agreement as difficulty; ties and near-miss values are not discussed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RadReason: Radiology Report Evaluation Metric with Reasons and Sub-Scores." pith.science (2026). https://pith.science/paper/PTMYVQCO

@misc{pith2026250815464,
  author       = {Pith},
  title        = {Pith review of: RadReason: Radiology Report Evaluation Metric with Reasons and Sub-Scores},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PTMYVQCO}},
  note         = {Machine review of arXiv:2508.15464}
}
read the original abstract

Evaluating automatically generated radiology reports remains a fundamental challenge due to the lack of clinically grounded, interpretable, and fine-grained metrics. Existing methods either produce coarse overall scores or rely on opaque black-box models, limiting their usefulness in real-world clinical workflows. We introduce RadReason, a novel evaluation framework for radiology reports that not only outputs fine-grained sub-scores across six clinically defined error types, but also produces human-readable justifications that explain the rationale behind each score. Our method builds on Group Relative Policy Optimization and incorporates two key innovations: (1) Sub-score Dynamic Weighting, which adaptively prioritizes clinically challenging error types based on live F1 statistics; and (2) Majority-Guided Advantage Scaling, which adjusts policy gradient updates based on prompt difficulty derived from sub-score agreement. Together, these components enable more stable optimization and better alignment with expert clinical judgment. Experiments on the ReXVal benchmark show that RadReason surpasses all prior offline metrics and achieves parity with GPT-4-based evaluations, while remaining explainable, cost-efficient, and suitable for clinical deployment. Code will be released upon publication.

Figures

Figures reproduced from arXiv: 2508.15464 by the authors.

Figure 1
Figure 1. Comparison of evaluation outputs across metric types. Traditional metrics yield only overall scores, lacking insight into specific errors. Our method provides both sub-scores and reasoning traces, enabling interpretable and detailed evaluation. Lavie, 2005; Lin, 2004), focus on word overlap and fail to capture clinically meaningful semantic dif￾ferences, particularly in cases involving paraphras￾ing, negation, or su… view at source ↗
Figure 2
Figure 2. Overview of our training framework. Left: The model produces detailed sub-scores and explanations across six clinically defined error types. Middle: During training, we apply Dynamic Sub-score Weighting by periodically computing F1 gaps and reweighting each dimension’s reward accordingly. Right: We introduce Majority-Guided Advantage Scaling, which estimates prompt difficulty by comparing majority-voted sub-scores a… view at source ↗
Figure 3
Figure 3. Case studies comparing our model’s sub-score predictions with human annotations, with step-by-step [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: GPT-4 Prompt Example. A Appendix A.1 GPT-4 Prompt Template The prompt in [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

32 extracted references · 11 canonical work pages

  1. [1]

    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 INTEGERS output.state before.all mid.sentence after.sentence after.block STRING...

  2. [2]

    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 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774

    arXiv 2023

  4. [4]

    Satanjeev Banerjee and Alon Lavie. 2005. Meteor: An automatic metric for mt evaluation with improved correlation with human judgments. In Proceedings of the acl workshop on intrinsic and extrinsic evaluation measures for machine translation and/or summarization, pages 65--72

    2005

  5. [5]

    Shruthi Bannur, Kenza Bouzid, Daniel C Castro, Anton Schwaighofer, Anja Thieme, Sam Bond-Taylor, Maximilian Ilse, Fernando P \'e rez-Garc \' a, Valentina Salvatelli, Harshita Sharma, and 1 others. 2024. Maira-2: Grounded radiology report generation. arXiv preprint arXiv:2406.04449

    arXiv 2024

  6. [6]

    Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadomski, Piotr Nyczyk, and 1 others. 2024. Graph of thoughts: Solving elaborate problems with large language models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 17682--17690

    work page 2024

  7. [7]

    Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783

    arXiv 2024

  8. [8]

    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

Show all 32 references
  1. [9]

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, and 1 others. 2022. Lora: Low-rank adaptation of large language models. ICLR, 1(2):3

    2022

  2. [10]

    Alyssa Huang, Oishi Banerjee, Kay Wu, Eduardo Pontes Reis, and Pranav Rajpurkar. 2024. Fineradscore: A radiology report line-by-line evaluation technique generating corrections with severity scores. In Machine Learning for Healthcare Conference. PMLR

    2024

  3. [11]

    Zhongzhen Huang, Xiaofan Zhang, and Shaoting Zhang. 2023. Kiut: Knowledge-injected u-transformer for radiology report generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 19809--19818

    2023

  4. [12]

    Saahil Jain, Ashwin Agrawal, Adriel Saporta, Steven QH Truong, Du Nguyen Duong, Tan Bui, Pierre Chambon, Yuhao Zhang, Matthew P Lungren, Andrew Y Ng, and 1 others. 2021. Radgraph: Extracting clinical entities and relations from radiology reports. arXiv preprint arXiv:2106.14463

    2021 arXiv

  5. [13]

    Komal Kumar, Tajamul Ashraf, Omkar Thawakar, Rao Muhammad Anwer, Hisham Cholakkal, Mubarak Shah, Ming-Hsuan Yang, Phillip HS Torr, Fahad Shahbaz Khan, and Salman Khan. 2025. Llm post-training: A deep dive into reasoning large language models. arXiv preprint arXiv:2502.21321

    2025 arXiv

  6. [14]

    Yingshu Li, Yunyi Liu, Zhanyu Wang, Xinyu Liang, Lingqiao Liu, Lei Wang, Leyang Cui, Zhaopeng Tu, Longyue Wang, and Luping Zhou. 2023. A comprehensive study of gpt-4v’s multimodal capabilities in medical imaging. medRxiv, pages 2023--11

    2023

  7. [15]

    Yingshu Li, Zhanyu Wang, Yunyi Liu, Lei Wang, Lingqiao Liu, and Luping Zhou. 2024. Kargen: Knowledge-enhanced automated radiology report generation using large language models. In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 382...

    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]

    Yunyi Liu, Zhanyu Wang, Yingshu Li, Xinyu Liang, Lingqiao Liu, Lei Wang, and Luping Zhou. 2024. Mrscore: Evaluating medical report with llm-based reward system. In International Conference on Medical Image Computing and Computer-Assisted Intervention, pages 283--292. Springer

    2024

  10. [18]

    Sophie Ostmeier, Justin Xu, Zhihong Chen, Maya Varma, Louis Blankemeier, Christian Bluethgen, Arne Md, Michael Moseley, Curtis Langlotz, Akshay Chaudhari, and 1 others. 2024. Green: Generative radiology report evaluation and error notation. In Findings of the Association for C...

    2024

  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]

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36:53728--53741

    2023

  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]

    Akshay Smit, Saahil Jain, Pranav Rajpurkar, Anuj Pareek, Andrew Y Ng, and Matthew P Lungren. 2020. Chexbert: combining automatic labelers and expert annotations for accurate radiology report labeling using bert. arXiv preprint arXiv:2004.09167

    2020 arXiv

  15. [23]

    Qwen Team. 2024. https://qwenlm.github.io/blog/qwen2.5/ Qwen2.5: A party of foundation models

    2024

  16. [24]

    Xiao Wang, Fuling Wang, Yuehang Li, Qingchuan Ma, Shiao Wang, Bo Jiang, Chuanfu Li, and Jin Tang. 2024. Cxpmrg-bench: Pre-training and benchmarking for x-ray medical report generation on chexpert plus dataset. arXiv preprint arXiv:2410.00379

    2024 arXiv

  17. [25]

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824--24837

    2022

  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]

    Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. Advances in neural information processing systems, 36:11809--11822

    2023

  20. [28]

    Feiyang Yu, Mark Endo, Rayan Krishnan, Ian Pan, Andy Tsai, Eduardo Pontes Reis, Eduardo Kaiser Ururahy Nunes Fonseca, Henrique Min Ho Lee, Zahra Shakeri Hossein Abad, Andrew Y Ng, and 1 others. 2023 a . Evaluating progress in automatic chest x-ray radiology report generation. ...

    2023

  21. [29]

    Feiyang Yu, Mark Endo, Rayan Krishnan, Ian Pan, Andy Tsai, Eduardo Pontes Reis, EKU Fonseca, Henrique Lee, Zahra Shakeri, Andrew Ng, and 1 others. 2023 b . Radiology report expert evaluation (rexval) dataset

    2023

  22. [30]

    Juan Manuel Zambrano Chaves, Shih-Cheng Huang, Yanbo Xu, Hanwen Xu, Naoto Usuyama, Sheng Zhang, Fei Wang, Yujia Xie, Mahmoud Khademi, Ziyi Yang, and 1 others. 2025. A clinically accessible small multimodal radiology model and evaluation metric for chest x-ray findings. Nature ...

    2025

  23. [31]

    Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q Weinberger, and Yoav Artzi. 2019. Bertscore: Evaluating text generation with bert. arXiv preprint arXiv:1904.09675

    2019 arXiv

  24. [32]

    Weike Zhao, Chaoyi Wu, Xiaoman Zhang, Ya Zhang, Yanfeng Wang, and Weidi Xie. 2024. Ratescore: A metric for radiology report generation. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 15004--15019

    2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.