Pith. sign in

REVIEW 2 major objections 8 minor 28 references

RadPhi-3: Small Language Models for Radiology

T0 review · 2 major / 8 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A 3.8B-parameter small language model, fine-tuned on a mixed radiology instruction corpus, is claimed to set a new state of the art on the RaLEs radiology report generation benchmark while also handling report section extraction and…

desk verdict RadPhi-3 is a solid incremental step beyond the authors' RadPhi-2, with useful new tasks and honest reporting, but the headline SOTA claim on RaLEs is not established because the MEDIQA-2021 references were GPT-4-cleaned only for RadPhi-3 and not for the baselines. read the letter →

arxiv 2411.13604 v1 pith:3WD3GWTO submitted 2024-11-19 cs.CV cs.CLcs.LG

Mercy Ranjit , Shaury Srivastav , Tanuja Ganu This is my paper · ORCID
classification cs.CVcs.CLcs.LG
keywords smalllanguagemodelsinstructiontuningradiologyreportgenerationchestX-rayimpressionpredictiontemporalchangesummarysectionsegmentationmultilinguallabels
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 argues that a small, locally deployable language model can carry a broad slice of radiology reporting work. The authors instruction-tune a 3.8B-parameter model on radiology question-answer pairs, chest X-ray report tasks, and two newly defined tasks—report section segmentation and temporal change summarization—and call the result RadPhi-3. They report that RadPhi-3 surpasses previously published results on the RaLEs radiology report generation benchmark, with the largest margins on the BioNLP-2023 dataset of CT and MRI reports from multiple anatomies. The payoff, if true, is that hospitals could get specialized radiology assistance without the cost, latency, and data-exposure risks of large API-hosted models.

What carries the argument

The load-bearing object is a single-stage instruction-tuning corpus built from public chest X-ray report annotations plus GPT-4-generated question-answer pairs and temporal change summaries. The backbone is Phi-3-mini-4k-instruct, an already instruction-tuned small language model, which the authors fine-tune for three epochs without a separate general-domain step or special token format. Prompt templates convert each source dataset into a task—impression prediction, abnormality and device labeling, QA comprehension, cleanup, extraction, section segmentation, temporal summaries, natural language inference, and Spanish label tagging—so one model is exposed to all tasks at once. That shared task mixture, together with clinical metrics such as F1-CheXbert and F1-RadGraph, is what the paper credits for the model's generalization.

What would settle it

Reproduce the RaLEs comparison with two changes: score RadPhi-3 on the original, uncleaned MEDIQA-2021 references, and apply the paper's cleanup prompt to the references used by the previously published baselines before computing their metrics; if the CheXbert or ROUGE-L gap reverses, the state-of-the-art claim is an artifact.

Watch

Extended reading notes

Core claim

RadPhi-3's central claim is that continued instruction tuning of Phi-3-mini-4k-instruct, a 3.8B-parameter small language model, produces a radiology system that matches or beats much larger systems on report generation and related tasks. On the RaLEs benchmark the paper reports an NLG score of 0.349 against the previous best 0.345, achieved through large gains on BioNLP-2023 (CheXbert 0.564 vs 0.506; ROUGE-L 0.310 vs 0.303) alongside mixed results on MEDIQA-2021, where lexical scores improve but CheXbert and RadGraph scores remain below the reported best. On radiology question answering, RadPhi-3 is reported to outperform both GPT-4 and the earlier RadPhi-2 in every stratified system, and it lifts natural language inference F1 from 0.17 (RadPhi-2) to 83.89. The paper also introduces two tasks—radiology report segmentation and temporal change summary—and reports high AlignScore and F1 on both, plus 97.64 macro F1 on Spanish abnormality-label prediction.

Load-bearing premise

The strongest claim stands on the fairness of the benchmark comparison: the MEDIQA-2021 reference impressions were cleaned with a GPT-4 prompt before scoring, and if the leaderboard baselines were not scored on equally cleaned references, the reported SOTA margin could be a preprocessing artifact rather than model superiority.

Editorial extensions

If this is right

  • If the benchmark claims hold, a 3.8B model can generate radiology impressions at or above the level of published larger systems, making local, on-premise deployment plausible for privacy-sensitive hospitals.
  • The BioNLP-2023 gains suggest the instruction tuning transfers beyond chest X-rays to CT and MRI reports of other anatomies, even though training used only chest X-ray reports.
  • The Padchest result indicates an English-trained model can label Spanish radiology reports with high F1, so multilingual extension may require little additional training.
  • Report section segmentation and temporal change summaries add workflow utilities that a single small model can provide alongside generation.
  • The NLI jump from 0.17 to 83.89 F1 over RadPhi-2 suggests that starting from an already instruction-tuned backbone fixes a semantic weakness of the previous small-model recipe.

Reading between the lines

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

  • The reported SOTA margin on MEDIQA-2021 should be treated cautiously: the paper cleaned that test set's reference impressions with a GPT-4 prompt before scoring, and if leaderboard baselines were scored on the original references, the comparison is not apples-to-apples.
  • Because the QA and change-summary training labels were generated by GPT-4 and audited on small samples (82 articles; 50 summaries), unknown label errors could inflate or deflate the reported quality; a larger radiologist audit is the natural check.
  • The strong BioNLP-2023 result hints that adding non-chest training data could yield further gains, an extension the paper leaves to future work.
  • A parameter-efficient low-rank fine-tune of the same backbone would be a cheap way to test how much of the benefit comes from the task mixture versus full-model continued pretraining; the paper lists this direction as future work.
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

2 major / 8 minor

Summary. The paper presents RadPhi-3, a 3.8B-parameter small language model instruction-tuned from Phi-3-mini-4k-instruct for a range of radiology tasks. The authors construct a large instruction-tuning dataset combining Radiopaedia-derived QA pairs (generated by GPT-4), MIMIC-CXR annotations, CheXpert Plus section annotations, and GPT-4-generated temporal change summaries. They evaluate the model on radiology QA, impression prediction, label prediction, section extraction, temporal change summarization, NLI, and the RaLEs benchmark, claiming state-of-the-art performance on RaLEs report generation. They also introduce two new tasks (radiology report segmentation and temporal change summary) and report generalization experiments on unseen tasks and on Spanish reports.

Significance. If the results hold, the paper demonstrates that a compact 3.8B model can match or exceed much larger models on targeted radiology tasks, which is valuable for privacy-sensitive and resource-constrained deployments. The use of clinical metrics (CheXbert, RadGraph), bootstrapped confidence intervals, and radiologist audits of GPT-4-generated datasets are methodological strengths that go beyond typical lexical-only evaluations. However, the headline SOTA claim on the RaLEs benchmark is based on an asymmetric evaluation, and the reported overall margin is small (0.349 vs 0.345). The practical significance of the work thus depends on whether the comparison can be put on a controlled footing.

major comments (2)
  1. [§4.2 and §4.9.1 (Tables 2 and 9)] The SOTA claims in these two sections rely on an uncontrolled comparison. In §4.2, the MIMIC-CXR test impressions are preprocessed with the cleanup radiology text prompt (Table 31) before computing metrics, while the Impression GPT baseline is cited from the original work and is presumably evaluated on the original references. In §4.9.1, the MEDIQA-2021 references are preprocessed for RadPhi-3 but the 'Best Reported' leaderboard numbers are not described as receiving the same preprocessing. This asymmetry gives RadPhi-3 a ROUGE-L advantage (+0.021 on MEDIQA-2021) while it scores lower on CheXbert (-0.028) and RadGraph (-0.053); the overall NLG margin is only 0.004, which is within the plausible range of the preprocessing effect. Please re-evaluate the existing baselines on the same preprocessed references, or report scores on the original references as well, and qualify the SOTA claims accordingly.
  2. [§3.1.1 and §3.1.4 (Tables 37 and 38)] The radiology QA test set and the temporal change summary test set are both generated by GPT-4. The reported radiologist audits (82 articles for QA, 50 summaries for temporal change) are small relative to the dataset sizes (93,068 QA pairs and 2,817 test summaries), and the temporal-summary audit found 5 category hallucinations and 7 missed findings. The QA evaluation in Table 1 is therefore an evaluation against model-generated references, which may favor a model trained on similarly generated data. Please provide a more detailed analysis of the potential test-set bias, or make the complete audit set publicly available for independent verification.
minor comments (8)
  1. [Abstract and §4.9.1] The phrase 'SOTA results on the RaLEs radiology report generation benchmark' should be qualified as 'for the NLG task' and should note that the overall margin is small and the comparison is affected by the reference-preprocessing issue.
  2. [Table 6] The table title 'RADLI + MEDNLI' is a typo; it should read 'RADNLI + MEDNLI'.
  3. [§3.4.1] The 'F1-score implementation of LlavaMed' is not a standard metric; please provide the exact version or a detailed description of how it is computed.
  4. [Table 9] The metric labels are inconsistent ('Rouge 2', 'RougeL', 'Chexbert', 'RG'); use standard abbreviations such as ROUGE-2, ROUGE-L, F1-CheXbert, and F1-RadGraph.
  5. [§4.9.1] The formula for the NLG score should be stated explicitly (e.g., the simple average of ROUGE-L and RadGraph across both datasets) rather than left implicit.
  6. [Throughout] There are several typos and formatting artifacts, including 'RadPhi-3is better thanRadPhi-2' (Table 1 caption), 'the the Phi-3-mini-4k-instruct' (§1), 'Mimic-CXR'/'MIMIC-CXR' inconsistency, 'spilt' (§3.1.2), and missing spaces between words.
  7. [§4.2] The paper states that MIMIC-CXR test impressions were preprocessed with GPT-4 but does not specify whether the RadPhi-2 baseline was evaluated on the same preprocessed test set; please clarify this for all reported comparisons.
  8. [Data availability] The statement 'We are not making the code available at this point' limits reproducibility; consider releasing at least the data-generation prompts and the evaluation scripts.

Circularity Check

1 steps flagged · score 4.0 of 10

The RaLEs SOTA claim rests on GPT-4-cleaned MEDIQA-2021 references used only for RadPhi-3; the overall margin is partly constructed, though the BIONLP-2023 result is independently meaningful.

  1. fitted input called prediction [Section 4.9.1, Table 9; Appendix E, Table 31]
    "We also note that as RadPhi-3 was trained on paraphrased versions of impressions of the MIMIC-CXR dataset, we preprocess the impression sections of the MEDIQA-2021 dataset which is also MIMIC-CXR based using the cleanup radiology text prompt defined in Table 31."

    The paper reports SOTA on the RaLEs benchmark by comparing with 'Best Reported' numbers in Table 9, but RadPhi-3's MEDIQA-2021 predictions are scored against impressions rewritten with the cleanup/paraphrase prompt, whose instruction is 'Paraphrase the findings into individual sentences without mention of priors.' The best-reported baselines are not described as receiving this rewrite. Since RadPhi-3 was instruction-tuned on paraphrased/cleaned impressions, the reference text is moved toward the model's training distribution. The overall NLG margin (0.349 vs 0.345) is therefore at least partly an artifact of the evaluation construction, not an independent derivation of superiority on the original benchmark.

full rationale

Most of the paper is not circular: RadPhi-3 is trained on public report annotations and evaluated on external benchmarks such as BIONLP-2023, Padchest, RadNLI/MEDNLI, and RadGraph2, and no parameter is fitted and then renamed as a prediction. The disclosed MEDIQA-2021 preprocessing is the one load-bearing construction: it changes the reference text for only the RadPhi-3 leg of the RaLEs comparison, partially manufacturing the overall SOTA margin. The BIONLP-2023 improvements (Rouge-L 0.310 vs 0.303; CheXbert 0.564 vs 0.506; RadGraph 0.325 vs 0.283) are independent and not affected by this issue. The GPT-4-generated QA and temporal-summary labels are a data-quality concern rather than a circular derivation; the small radiologist audits provide partial external support and the central claim does not reduce to a self-citation chain.

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

The central claims rest on GPT-4-generated training and evaluation labels for two of the main tasks, on existing public annotations reused as ground truth, and on standard fine-tuning hyperparameters chosen without reported ablations. No new physical or mathematical entities are postulated; the paper's contribution is an engineered model, so the ledger mostly records dataset and metric assumptions.

free parameters (4)
  • learning_rate = 5e-5
    Training hyperparameter chosen by hand for the fine-tuning run (Section 3.2); affects final model quality but is not fitted to a benchmark.
  • global_batch_size = 128
    Chosen for 3-epoch fine-tuning on 4x A100 GPUs (Section 3.2).
  • num_epochs = 3
    Fine-tuning duration chosen without reported ablation (Section 3.2).
  • warmup_ratio = 0.1
    Standard warmup fraction (Section 3.2).
assumptions (4)
  • domain assumption GPT-4-generated QA pairs from Radiopaedia articles are factually correct and clinically useful as ground truth.
    Used to build both train and test QA sets (Section 3.1.1). A radiologist audit of 82 of 15,076 articles found zero factual hallucinations but 5 quality errors and 1 instruction-following error, so the assumption is partially validated, not fully.
  • domain assumption CheXpert Plus sections and RadGraph2/MIMIC-CXR change annotations are accurate ground truth for segmentation and temporal summary tasks.
    The model is trained and evaluated directly against these existing annotations (Sections 3.1.3, 3.1.4); no independent re-labeling is done for the full dataset.
  • domain assumption Phi-3-mini-4k-instruct is a suitable base model whose pretrained knowledge transfers to radiology.
    The authors fine-tune this specific checkpoint (Section 3.2) and attribute NLI improvements to its instruction-tuned nature, but do not ablate over other small bases.
  • domain assumption AlignScore and F1-CheXbert are valid proxies for factual consistency and clinical correctness.
    These metrics are used to support claims of factuality and clinical quality (Sections 3.4.2, 3.4.3); the paper does not validate them against human judgment on the model outputs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RadPhi-3: Small Language Models for Radiology." pith.science (2026). https://pith.science/paper/3WD3GWTO

@misc{pith2026241113604,
  author       = {Pith},
  title        = {Pith review of: RadPhi-3: Small Language Models for Radiology},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3WD3GWTO}},
  note         = {Machine review of arXiv:2411.13604}
}
read the original abstract

LLM based copilot assistants are useful in everyday tasks. There is a proliferation in the exploration of AI assistant use cases to support radiology workflows in a reliable manner. In this work, we present RadPhi-3, a Small Language Model instruction tuned from Phi-3-mini-4k-instruct with 3.8B parameters to assist with various tasks in radiology workflows. While impression summary generation has been the primary task which has been explored in prior works w.r.t radiology reports of Chest X-rays, we also explore other useful tasks like change summary generation comparing the current radiology report and its prior report, section extraction from radiology reports, tagging the reports with various pathologies and tubes, lines or devices present in them etc. In-addition, instruction tuning RadPhi-3 involved learning from a credible knowledge source used by radiologists, Radiopaedia.org. RadPhi-3 can be used both to give reliable answers for radiology related queries as well as perform useful tasks related to radiology reports. RadPhi-3 achieves SOTA results on the RaLEs radiology report generation benchmark.

Figures

Figures reproduced from arXiv: 2411.13604 by the authors.

Figure 1
Figure 1. Training Workflow of RadPhi-3. observations and support devices. This is extracted under 5 different categories (New, Resolved, Wors￾ened, Stable and Improved) for medical conditions and 4 different categories (New, Removed, Changed and Unchanged) for support devices. In addition, we also had a category for extracting the negative findings for medical conditions and rec￾ommendations for device adjustments. We use GP… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 11 canonical work pages

  1. [1]

    Marah Abdin, Jyoti Aneja, Hany Awadalla, Ahmed Awadallah, Ammar Ahmad Awan, Nguyen Bach, Amit Bahree, Arash Bakhtiari, Jianmin Bao, Harkirat Behl, Alon Benhaim, Misha Bilenko, Johan Bjorck, Sébastien Bubeck, Martin Cai, Qin Cai, Vishrav Chaudhary, Dong Chen, Dongdong Chen, Weizhu Chen, Yen-Chun Chen, Yi-Ling Chen, Hao Cheng, Parul Chopra, Xiyang Dai, Matt...

    arXiv 2024

  2. [2]

    MS-CXR-T : Learning to exploit temporal structure for biomedical vision-language processing, 2023

    Shruthi Bannur, Stephanie Hyland, Qianchu Liu, Fernando P \'e rez-Garc \' a, Max Ilse, Daniel Coelho de Castro, Benedikt Boecking, Harshita Sharma, Kenza Bouzid, Anton Schwaighofer, Maria Teodora Wetscherek, Hannah Richardson, Tristan Naumann, Javier Alvarez Valle, and Ozan Oktay. MS-CXR-T : Learning to exploit temporal structure for biomedical vision-lan...

    work page 2023

  3. [3]

    Overview of the MEDIQA 2021 shared task on summarization in the medical domain

    Asma Ben Abacha, Yassine Mrabet, Yuhao Zhang, Chaitanya Shivade, Curtis Langlotz, and Dina Demner-Fushman. Overview of the MEDIQA 2021 shared task on summarization in the medical domain. In Dina Demner-Fushman, Kevin Bretonnel Cohen, Sophia Ananiadou, and Junichi Tsujii, editors, Proceedings of the 20th Workshop on Biomedical Language Processing, pages 74...

    doi:10.18653/v1/2021.bionlp-1.8 2021

  4. [4]

    Padchest: A large chest x-ray image dataset with multi-label annotated reports, 2019

    Aurelia Bustos, Antonio Pertusa, Jose-Maria Salinas, and Maria de la Iglesia-Vayá. Padchest: A large chest x-ray image dataset with multi-label annotated reports, 2019. URL https://arxiv.org/abs/1901.07441

    arXiv 2019

  5. [5]

    Langlotz

    Pierre Chambon, Jean-Benoit Delbrouck, Thomas Sounack, Shih-Cheng Huang, Zhihong Chen, Maya Varma, Steven QH Truong, Chu The Chuong, and Curtis P. Langlotz. Chexpert plus: Augmenting a large chest x-ray dataset with text radiology reports, patient demographics and additional image formats, 2024. URL https://arxiv.org/abs/2405.19538

    arXiv 2024

  6. [6]

    Ra LE s: a benchmark for radiology language evaluations

    Juan Manuel Zambrano Chaves, Nandita Bhaskhar, Maayane Attias, Jean-Benoit Delbrouck, Daniel Rubin, Andreas Markus Loening, Curtis Langlotz, and Akshay S Chaudhari. Ra LE s: a benchmark for radiology language evaluations. In Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track, 2023. URL https://openreview.net/f...

    work page 2023

  7. [7]

    RadGraph2 : Tracking findings over time in radiology reports, 2024

    Adam Dejl, Sameer Khanna, Patricia Therese Pile, Kibo Yoon, Steven Q H Truong, Hanh Duong, Agustina Saenz, and Pranav Rajpurkar. RadGraph2 : Tracking findings over time in radiology reports, 2024

    work page 2024

  8. [8]

    Overview of the R ad S um23 shared task on multi-modal and multi-anatomical radiology report summarization

    Jean-Benoit Delbrouck, Maya Varma, Pierre Chambon, and Curtis Langlotz. Overview of the R ad S um23 shared task on multi-modal and multi-anatomical radiology report summarization. In Dina Demner-fushman, Sophia Ananiadou, and Kevin Cohen, editors, The 22nd Workshop on Biomedical Natural Language Processing and BioNLP Shared Tasks, pages 478--482, Toronto,...

    doi:10.18653/v1/2023.bionlp-1.45 2023

Show all 28 references
  1. [9]

    Summers, and Yingying Zhu

    Xinyue Hu, Lin Gu, Qiyuan An, Mengliang Zhang, Liangchen Liu, Kazuma Kobayashi, Tatsuya Harada, Ronald M. Summers, and Yingying Zhu. Expert knowledge-aware image difference graph representation learning for difference-aware medical visual question answering. In Proceedings of ...

    2023

  2. [10]

    Lungren, Andrew Y

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

    2021 arXiv

  3. [11]

    Alistair E. W. Johnson, Tom J. Pollard, Seth J. Berkowitz, Nathaniel R. Greenbaum, Matthew P. Lungren, Chih-ying Deng, Roger G. Mark, and Steven Horng. Mimic-cxr, a de-identified publicly available database of chest radiographs with free-text reports. Scientific Data, 6 0 (1):...

    2019 doi

  4. [12]

    Radgraph2: Modeling disease progression in radiology reports via hierarchical information extraction

    Sameer Khanna, Adam Dejl, Kibo Yoon, Steven QH Truong, Hanh Duong, Agustina Saenz, and Pranav Rajpurkar. Radgraph2: Modeling disease progression in radiology reports via hierarchical information extraction. In Kaivalya Deshpande, Madalina Fiterau, Shalmali Joshi, Zachary Lipto...

    2023

  5. [13]

    Llava-med: Training a large language-and-vision assistant for biomedicine in one day, 2023

    Chunyuan Li, Cliff Wong, Sheng Zhang, Naoto Usuyama, Haotian Liu, Jianwei Yang, Tristan Naumann, Hoifung Poon, and Jianfeng Gao. Llava-med: Training a large language-and-vision assistant for biomedicine in one day, 2023. URL https://arxiv.org/abs/2306.00890

    2023 arXiv

  6. [14]

    ROUGE : A package for automatic evaluation of summaries

    Chin-Yew Lin. ROUGE : A package for automatic evaluation of summaries. In Text Summarization Branches Out, pages 74--81, Barcelona, Spain, July 2004. Association for Computational Linguistics. URL https://aclanthology.org/W04-1013

    2004

  7. [15]

    Exploring the boundaries of GPT -4 in radiology

    Qianchu Liu, Stephanie Hyland, Shruthi Bannur, Kenza Bouzid, Daniel Castro, Maria Wetscherek, Robert Tinn, Harshita Sharma, Fernando P \'e rez-Garc \' a, Anton Schwaighofer, Pranav Rajpurkar, Sameer Khanna, Hoifung Poon, Naoto Usuyama, Anja Thieme, Aditya Nori, Matthew Lungren...

    2023

  8. [16]

    Radiology-llama2: Best-in-class large language model for radiology, 2023 b

    Zhengliang Liu, Yiwei Li, Peng Shu, Aoxiao Zhong, Longtao Yang, Chao Ju, Zihao Wu, Chong Ma, Jie Luo, Cheng Chen, Sekeun Kim, Jiang Hu, Haixing Dai, Lin Zhao, Dajiang Zhu, Jun Liu, Wei Liu, Dinggang Shen, Tianming Liu, Quanzheng Li, and Xiang Li. Radiology-llama2: Best-in-clas...

    2023 arXiv

  9. [17]

    Radiology-gpt: A large language model for radiology, 2024

    Zhengliang Liu, Aoxiao Zhong, Yiwei Li, Longtao Yang, Chao Ju, Zihao Wu, Chong Ma, Peng Shu, Cheng Chen, Sekeun Kim, Haixing Dai, Lin Zhao, Lichao Sun, Dajiang Zhu, Jun Liu, Wei Liu, Dinggang Shen, Xiang Li, Quanzheng Li, and Tianming Liu. Radiology-gpt: A large language model...

    2024 arXiv

  10. [18]

    An iterative optimizing framework for radiology report summarization with chatgpt

    Chong Ma, Zihao Wu, Jiaqi Wang, Shaochen Xu, Yaonai Wei, Zhengliang Liu, Fang Zeng, Xi Jiang, Lei Guo, Xiaoyan Cai, Shu Zhang, Tuo Zhang, Dajiang Zhu, Dinggang Shen, Tianming Liu, and Xiang Li. An iterative optimizing framework for radiology report summarization with chatgpt. ...

    2024

  11. [19]

    RadNLI : A natural language inference dataset for the radiology domain, 2021

    Yasuhide Miura, Yuhao Zhang, Emily Tsai, Curtis Langlotz, and Dan Jurafsky. RadNLI : A natural language inference dataset for the radiology domain, 2021

    2021

  12. [20]

    Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, and...

    2022 arXiv

  13. [21]

    B leu: a method for automatic evaluation of machine translation

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. B leu: a method for automatic evaluation of machine translation. In Pierre Isabelle, Eugene Charniak, and Dekang Lin, editors, Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, ...

    2002

  14. [22]

    Rad-Phi2 : Instruction tuning phi2 for radiology

    Mercy Ranjit, Gopinath Ganapathy, Shaury Srivastav, Srujana Oruganti, and Tanuja Ganu. Rad-Phi2 : Instruction tuning phi2 for radiology. Advances in Artificial Intelligence and Machine Learning, 04 0 (02): 0 2302--2323, 2024

    2024

  15. [23]

    MedNLI --- a natural language inference dataset for the clinical domain, 2017

    Chaitanya Shivade. MedNLI --- a natural language inference dataset for the clinical domain, 2017

    2017

  16. [24]

    Combining automatic labelers and expert annotations for accurate radiology report labeling using BERT

    Akshay Smit, Saahil Jain, Pranav Rajpurkar, Anuj Pareek, Andrew Ng, and Matthew Lungren. Combining automatic labelers and expert annotations for accurate radiology report labeling using BERT . In Bonnie Webber, Trevor Cohn, Yulan He, and Yang Liu, editors, Proceedings of the 2...

    2020 doi

  17. [25]

    RadQA : A question answering dataset to improve comprehension of radiology reports, 2022

    Sarvesh Soni and Kirk Roberts. RadQA : A question answering dataset to improve comprehension of radiology reports, 2022

    2022

  18. [26]

    Chest ImaGenome dataset, 2021

    Joy Wu, Nkechinyere Agu, Ismini Lourentzou, Arjun Sharma, Joseph Paguio, Jasper Seth Yao, Edward Christopher Dee, William Mitchell, Satyananda Kashyap, Andrea Giovannini, Leo Anthony Celi, Tanveer Syeda-Mahmood, and Mehdi Moradi. Chest ImaGenome dataset, 2021

    2021

  19. [27]

    Alignscore: Evaluating factual consistency with a unified alignment function, 2023

    Yuheng Zha, Yichi Yang, Ruichen Li, and Zhiting Hu. Alignscore: Evaluating factual consistency with a unified alignment function, 2023. URL https://arxiv.org/abs/2305.16739

    2023 arXiv

  20. [28]

    Long-tailed extreme multi-label text classification by the retrieval of generated pseudo label descriptions

    Ruohong Zhang, Yau-Shian Wang, Yiming Yang, Donghan Yu, Tom Vu, and Likun Lei. Long-tailed extreme multi-label text classification by the retrieval of generated pseudo label descriptions. In Andreas Vlachos and Isabelle Augenstein, editors, Findings of the Association for Comp...

    2023 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.