High-Fidelity Pseudo-label Generation by Large Language Models for Training Robust Radiology Report Classifiers

Brian Wong; Kaito Tanaka

REVIEW 3 major objections 5 minor 30 references

High-Fidelity Pseudo-label Generation by Large Language Models for Training Robust Radiology Report Classifiers

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 DeBERTa-RAD, a compact model distilled from LLM pseudo-labels, reaches a Macro F1 of 0.9120 on the MIMIC-500 chest X-ray report labeling benchmark, outperforming rule-based, supervised, and direct-LLM baselines.

desk verdict A new MIMIC-500 F1 number from an LLM-pseudo-label pipeline, but the distillation mechanism is mis-specified and the train/test split is unstated—needs major revision. read the letter →

arxiv 2505.01693 v1 pith:4DJGXJAY submitted 2025-05-03 cs.CL

Brian Wong , Kaito Tanaka This is my paper · ORCID

classification cs.CL

keywords radiologyreportlabelingknowledgedistillationpseudo-labelingDeBERTachestX-rayuncertaintyclassificationmedicalNLPMIMIC-CXR

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

Automated labeling of chest X-ray reports into Present, Absent, or Uncertain for 13 findings is hard because clinical prose is full of negation and hedging. This paper argues that large-scale human annotation is unnecessary: a large language model can generate pseudo-labels for over 200,000 reports, and a small DeBERTa-Base model trained on those labels via knowledge distillation can match or beat expert-level baseline systems. The paper reports a Macro F1 of 0.9120 on the expert-annotated MIMIC-500 test set, above CheXpert, CheXbert, GPT-4 direct inference, and the prior CheX-GPT approach, with inference around 750 reports per second. The claimed significance is that accurate medical text extraction can be obtained cheaply and quickly without manual annotation.

What carries the argument

The machinery is a two-stage teacher-student setup. An advanced LLM (GPT-4) reads each report with a prompt that asks for the status (Present, Absent, Uncertain) of each of 13 findings, producing pseudo-labels for over 200,000 reports. The student is DeBERTa-Base (about 86 million parameters), whose disentangled attention is expected to capture long-range dependencies in clinical sentences; one classification head per finding maps the [CLS] representation to a three-way softmax. The total training loss combines a hard-target cross-entropy term with a temperature-scaled distillation term, both supervised by the LLM's hard pseudo-labels.

What would settle it

Check the overlap between the MIMIC-500 test set and the pseudo-labeled training corpus D_pseudo; if any test report appears in training, re-run the evaluation on the subset of test reports that have no training overlap and compare Macro F1.

Watch

Extended reading notes

Core claim

The central claim is that a two-stage distillation pipeline, LLM pseudo-labeling followed by training a DeBERTa-Base student, produces a radiology report labeler that is simultaneously more accurate and much faster than direct LLM inference. On the MIMIC-500 benchmark the student reaches a Macro F1 of 0.9120 across 13 findings with three-way certainty status, statistically significantly above the rule-based CheXpert (0.8864), the supervised CheXbert (0.9047), GPT-4 direct inference (0.9014), and CheX-GPT (0.9014). The paper further claims the largest relative gain occurs on findings marked Uncertain (F1 0.852 vs. 0.798 for CheXbert), and that expert radiologists judged the student's predictions more accurate than CheXbert's in 63.7% of disagreement cases.

Load-bearing premise

The load-bearing premise is that the 500 expert-annotated MIMIC-500 test reports are not among the over-200,000 reports used to generate the pseudo-labeled training corpus, so the reported F1 measures generalization rather than memorization.

Editorial extensions

If this is right

If the result holds, a small model can replace direct LLM APIs for large-scale chest X-ray labeling, reducing per-report cost and latency by several orders of magnitude.
The uncertain-status gain (0.852 vs. 0.798) implies that distilled pseudo-labels preserve clinically important hedging cues that rule-based and supervised systems miss.
The framework removes the need for large manually annotated corpora; the same two-stage recipe could be applied to other structured extraction tasks in clinical text.
Statistical significance over all baselines (p < 0.05) strengthens the claim that the gap is not noise, assuming the evaluation split is clean.

Reading between the lines

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

If the MIMIC-500 test reports overlap with the pseudo-labeled training corpus, the headline F1 gain could be partly memorization; a clean-split replication is the natural next check.
Because the teacher only emits hard labels, the distillation term as written is closer to label smoothing than to classic soft-target distillation; a teacher that emits probability distributions might improve the uncertain class further.
The same pseudo-label-plus-distillation recipe is a testable extension for other report types (e.g., MRI, pathology) and other label taxonomies.
One could measure pseudo-label quality per finding to see which findings the LLM teacher gets wrong and whether student errors mirror those teacher errors.

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.

Desk Editor's Note

read the letter

DeBERTa-RAD is a two-stage pipeline—GPT-4 pseudo-labels on ~200k MIMIC-CXR reports, then train a DeBERTa-Base classifier—and the reported 0.9120 Macro F1 on MIMIC-500 is a new number. The paper does several things right: it evaluates against the relevant baselines (CheXpert, CheXbert, GPT-4 direct, CheX-GPT), reports per-finding and per-certainty breakdowns, and includes a human evaluation and a pseudo-label quality check. The uncertainty analysis is the most interesting part: an 0.852 F1 on Uncertain labels, up from 0.798 for CheXbert.

But the central mechanistic claim doesn't survive contact with the equations. The 'tailored knowledge distillation' loss is cross-entropy against a one-hot teacher distribution with a temperature-scaled student softmax. Eq. (5)–(8) show it's the same loss as L_hard with a temperature parameter; there is no soft teacher distribution, so it cannot transfer confidence or decision margins. The paper attributes the uncertainty gains to this mechanism, which the math does not support. That's a load-bearing error, not a cosmetic one.

The bigger issue for the headline number: Section IV-A says D_pseudo is 'a large subset of the MIMIC-CXR reports' and never states that the 500 MIMIC-500 test reports were removed. Since MIMIC-500 is a subset of MIMIC-CXR, the evaluation set may be inside the training corpus. If so, the 0.9120 is partly memorization, and the comparisons against CheXbert and CheX-GPT are not fair. A one-sentence exclusion statement would fix this, but its absence makes the main result unverifiable. Combined with missing code and unreported hyperparameters, the reported gain of 0.7 points over CheXbert is hard to take at face value.

This is not a circularity in the sense of fitting to the test labels—the pseudo-labels are external and the gold labels are human—but the evaluation integrity is unconfirmed. The paper deserves a serious referee, not because the claims are credible as written, but because the flaws are concrete and addressable. I'd ask for a major revision: state the train/test split explicitly, release code and hyperparameters, and either correct the distillation description or drop the KD framing. If the authors can do that, the empirical result could be a useful incremental contribution. As written, I would not accept.

Referee Report

3 major / 5 minor

Summary. The paper proposes DeBERTa-RAD, a two-stage framework for chest X-ray report labeling. In the first stage, an advanced LLM generates pseudo-labels (Present, Absent, Uncertain) for 13 findings across a large corpus of MIMIC-CXR reports. In the second stage, a DeBERTa-Base student model is trained on these pseudo-labels using a loss that combines a hard-label cross-entropy term with a temperature-scaled cross-entropy term, which the authors describe as knowledge distillation. The student model is evaluated on the expert-annotated MIMIC-500 test set, where it reports a Macro F1 of 0.9120, claiming state-of-the-art performance over CheXpert, CheXbert, GPT-4 direct inference, and CheX-GPT, along with better handling of uncertain findings.

Significance. If the evaluation is uncontaminated and the method's mechanism is correctly described, the paper would offer a practical solution for scalable radiology report labeling without human annotation, and the reported gains on uncertain findings would be clinically meaningful. However, two load-bearing issues affect the validity of the central claim: a possible train/test leakage because the training corpus is drawn from the same dataset as the test set without an explicit exclusion statement, and a distillation loss that is mathematically equivalent to the hard-label loss because the teacher distribution is one-hot. The manuscript also leaves the statistical testing procedure and several evaluation details underspecified.

major comments (3)

[Section IV-A, Table I] The manuscript does not state that the 500 MIMIC-500 test reports were excluded from D_pseudo, which is described as 'a large subset of the MIMIC-CXR reports.' Since MIMIC-500 is a subset of MIMIC-CXR, the training corpus may contain the exact reports used for evaluation. If this is the case, the comparisons in Table I and the significance tests in Table II are invalid: DeBERTa-RAD could memorize the test reports, while baselines such as CheXbert were trained on external CheXpert labels. Please state explicitly whether and how the test set was held out; if it was not excluded, the evaluation must be re-run on a disjoint split.
[Section III-C, Eqs. (4)-(7)] The distillation loss L_distill uses the one-hot teacher distribution P_hard_teacher. With this definition, L_distill is mathematically identical to L_hard except for the temperature in the student's softmax: both are negative log-likelihoods of the same hard pseudo-label. The claim that this term 'encourages the student to learn the relative probabilities implied by the teacher's decision, even if the teacher's output is just a hard label' is therefore not supported, because a one-hot target carries no relative probabilities or confidence. Consequently, the paper's attribution of improved uncertainty handling to knowledge distillation (Sections IV-E and V) is unsubstantiated; the actual method is a weighted combination of two cross-entropy losses with identical targets.
[Section IV-D, Table II] The statistical significance testing is underspecified. The text states that 'paired t-tests and Wilcoxon signed-rank tests comparing the F1 scores per finding category' were conducted 'across the 500 test reports,' but F1 per finding is a scalar per model, not a distribution over reports. It is unclear what the unit of analysis was (report-level predictions per finding, per-report F1, or per-finding F1), and whether the 13 findings were treated as independent samples. Please describe the exact test statistic, the unit of analysis, and whether any multiple-comparison correction was applied; the p-values in Table II are not interpretable without this information.

minor comments (5)

[Section III-A] The mapping of 'Not Mentioned' pseudo-labels to the three-class prediction target is not defined; please clarify whether reports with no mention of a finding are excluded from the loss, treated as Absent, or handled by a fourth class.
[Table I] CheX-GPT and GPT-4 Direct Inference are reported with exactly the same Macro F1 (0.9014); please confirm this is not a typographical error and explain how two different systems could produce identical scores.
[Section IV-F, Table III] The human evaluation lacks essential methodological details: the number of radiologist annotators, the instructions they received, inter-rater reliability, and confidence intervals for the reported percentages; please add these.
[Section IV-D, first paragraph] The reference to 'Section II' is incorrect; the loss formulation is described in Section III, not Section II.
[References] Reference [29] is identical to reference [25]; they should be merged or one should be removed.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the headline F1 is an empirical benchmark result against independent human labels; the redundant distillation loss and the unstated MIMIC-500/D_pseudo overlap are methodological concerns, not circular derivations.

full rationale

DeBERTa-RAD's core result is not circular: pseudo-labels are generated by an external LLM (GPT-4), the student is trained on a large MIMIC-CXR corpus, and MIMIC-500 human expert labels are used only for evaluation. No parameter is fitted to the gold test labels, so the reported Macro F1 is an empirical outcome rather than a consequence of the method's definitions. I considered the distillation loss: the equations show L_distill is cross-entropy against the same one-hot pseudo-label as L_hard, with only the student temperature changed; this makes the claimed transfer of the teacher's internal margin vacuous, but it is a conflation or redundancy in the mechanism narrative, not a circular derivation of the benchmark numbers. I also considered the unstated relationship between D_pseudo and MIMIC-500: since D_pseudo is only described as 'a large subset of the MIMIC-CXR reports,' the paper never proves the 500 test reports were excluded; if they were included, the comparison would be contaminated, but contamination is a data-leakage or correctness risk rather than circularity under the definitions used here. Citations to DeBERTa and weak-to-strong generalization are external references, not self-citations of the authors, and no uniqueness theorem is invoked. Overall, the main claim remains independently testable against human gold labels, so the circularity score is low.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The method rests on two empirical premises, teacher label quality and test-set disjointness, plus one conceptual premise about what counts as distillation. The first is partially measured, the second is unstated, and the third is, by the paper's own equations, not a form of knowledge transfer.

free parameters (3)

alpha (loss weighting) = not reported
Weight between L_distill and L_hard in Eq. (8); tuned on a validation subset of the pseudo-labeled data (Section IV-C).
T_distill (temperature) = not reported
Softmax temperature for the distillation term; tuned on a validation subset (Section IV-C).
learning rate, batch size, epochs = not reported
Training hyperparameters tuned on a validation subset; final values not given (Section IV-C).

assumptions (3)

domain assumption GPT-4 generates pseudo-labels of sufficient quality and consistency to train a strong student model.
The method's first stage depends entirely on the teacher's label quality; Section IV-I reports 0.9014 Macro F1 versus human gold on MIMIC-500, and any teacher errors are inherited by the student.
domain assumption The MIMIC-500 test set is disjoint from the pseudo-labeled training corpus D_pseudo.
Section IV-A describes D_pseudo as a large subset of MIMIC-CXR but never states that the 500 evaluation reports were excluded, so a clean benchmark comparison is assumed.
ad hoc to paper A one-hot hard label combined with a temperature-scaled student softmax constitutes knowledge distillation.
Section III-C Eq. (5)-(7) defines the teacher distribution as one-hot, which carries no confidence information; this assumption is required for the paper's central framing but is unjustified.

how reviews work

0 comments

Cite this review

Pith. "Pith review of High-Fidelity Pseudo-label Generation by Large Language Models for Training Robust Radiology Report Classifiers." pith.science (2026). https://pith.science/paper/4DJGXJAY

@misc{pith2026250501693,
  author       = {Pith},
  title        = {Pith review of: High-Fidelity Pseudo-label Generation by Large Language Models for Training Robust Radiology Report Classifiers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4DJGXJAY}},
  note         = {Machine review of arXiv:2505.01693}
}

read the original abstract

Automated labeling of chest X-ray reports is essential for enabling downstream tasks such as training image-based diagnostic models, population health studies, and clinical decision support. However, the high variability, complexity, and prevalence of negation and uncertainty in these free-text reports pose significant challenges for traditional Natural Language Processing methods. While large language models (LLMs) demonstrate strong text understanding, their direct application for large-scale, efficient labeling is limited by computational cost and speed. This paper introduces DeBERTa-RAD, a novel two-stage framework that combines the power of state-of-the-art LLM pseudo-labeling with efficient DeBERTa-based knowledge distillation for accurate and fast chest X-ray report labeling. We leverage an advanced LLM to generate high-quality pseudo-labels, including certainty statuses, for a large corpus of reports. Subsequently, a DeBERTa-Base model is trained on this pseudo-labeled data using a tailored knowledge distillation strategy. Evaluated on the expert-annotated MIMIC-500 benchmark, DeBERTa-RAD achieves a state-of-the-art Macro F1 score of 0.9120, significantly outperforming established rule-based systems, fine-tuned transformer models, and direct LLM inference, while maintaining a practical inference speed suitable for high-throughput applications. Our analysis shows particular strength in handling uncertain findings. This work demonstrates a promising path to overcome data annotation bottlenecks and achieve high-performance medical text processing through the strategic combination of LLM capabilities and efficient student models trained via distillation.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 18 canonical work pages

[1]

Mimic-cxr, a de- identified publicly available database of chest radiographs with free-text reports,

A. E. Johnson, T. J. Pollard, S. J. Berkowitz, N. R. Greenbaum, M. P. Lungren, C.-y. Deng, R. G. Mark, and S. Horng, “Mimic-cxr, a de- identified publicly available database of chest radiographs with free-text reports,”Scientific data, vol. 6, no. 1, p. 317

work page
[2]

Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison,

J. Irvin, P. Rajpurkar, M. Ko, Y . Yu, S. Ciurea-Ilcus, C. Chute, H. Mark- lund, B. Haghgoo, R. Ball, K. Shpanskayaet al., “Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison,” inProceedings of the AAAI conference on artificial intelligence, vol. 33, no. 01, pp. 590–597

work page
[3]

Enhancing intent understanding for ambiguous prompts through human-machine co-adaptation,

Y . He, J. Wang, K. Li, Y . Wang, L. Sun, J. Yin, M. Zhang, and X. Wang, “Enhancing intent understanding for ambiguous prompts through human-machine co-adaptation,”arXiv preprint arXiv:2501.15167, 2025

arXiv 2025
[4]

Natural language processing in radiology: a systematic review,

E. Pons, L. M. Braun, M. M. Hunink, and J. A. Kors, “Natural language processing in radiology: a systematic review,”Radiology, vol. 279, no. 2, pp. 329–343

work page
[5]

Chexbert: combining automatic labelers and expert annota- tions for accurate radiology report labeling using bert,

A. Smit, S. Jain, P. Rajpurkar, A. Pareek, A. Y . Ng, and M. P. Lungren, “Chexbert: combining automatic labelers and expert annota- tions for accurate radiology report labeling using bert,”arXiv preprint arXiv:2004.09167

arXiv 2004
[6]

A systematic review of natural language processing applied to radiology reports,

A. Casey, E. Davidson, M. Poon, H. Dong, D. Duma, A. Grivas, C. Grover, V . Su ´arez-Paniagua, R. Tobin, W. Whiteleyet al., “A systematic review of natural language processing applied to radiology reports,”BMC medical informatics and decision making, vol. 21, no. 1, p. 179

work page
[7]

Utilizing bert for biomedical and clinical text mining,

R. Zhu, X. Tu, and J. X. Huang, “Utilizing bert for biomedical and clinical text mining,” inData analytics in biomedical engineering and healthcare. Elsevier, pp. 73–103

work page
[8]

Thread of thought unraveling chaotic contexts,

Y . Zhou, X. Geng, T. Shen, C. Tao, G. Long, J.-G. Lou, and J. Shen, “Thread of thought unraveling chaotic contexts,”arXiv preprint arXiv:2311.08734, 2023

arXiv 2023

Show all 30 references

[9]

Less is more: Vision representation compression for efficient video generation with large language models,

Y . Zhou, J. Zhang, G. Chen, J. Shen, and Y . Cheng, “Less is more: Vision representation compression for efficient video generation with large language models,” 2024

2024
[10]

Deberta: Decoding-enhanced bert with disentangled attention,

P. He, X. Liu, J. Gao, and W. Chen, “Deberta: Decoding-enhanced bert with disentangled attention,”arXiv preprint arXiv:2006.03654

2006 arXiv
[11]

Weak to strong generalization for large language models with multi-capabilities,

Y . Zhou, J. Shen, and Y . Cheng, “Weak to strong generalization for large language models with multi-capabilities,” inThe Thirteenth International Conference on Learning Representations, 2025. [Online]. Available: https://openreview.net/forum?id=N1vYivuSKq 8

2025
[12]

Attention is all you need,

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,”Advances in neural information processing systems, vol. 30
[13]

Bert: Pre-training of deep bidirectional transformers for language understanding,

J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” inPro- ceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technolog...

2019
[14]

Language mod- els are few-shot learners,

T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askellet al., “Language mod- els are few-shot learners,”Advances in neural information processing systems, vol. 33, pp. 1877–1901

1901
[15]

Scaling laws for neural language models,

J. Kaplan, S. McCandlish, T. Henighan, T. B. Brown, B. Chess, R. Child, S. Gray, A. Radford, J. Wu, and D. Amodei, “Scaling laws for neural language models,”arXiv preprint arXiv:2001.08361

2001 arXiv
[16]

Lamda: Language models for dialog applications,

R. Thoppilan, D. De Freitas, J. Hall, N. Shazeer, A. Kulshreshtha, H.-T. Cheng, A. Jin, T. Bos, L. Baker, Y . Duet al., “Lamda: Language models for dialog applications,”arXiv preprint arXiv:2201.08239

arXiv
[17]

Palm: Scal- ing language modeling with pathways,

A. Chowdhery, S. Narang, J. Devlin, M. Bosma, G. Mishra, A. Roberts, P. Barham, H. W. Chung, C. Sutton, S. Gehrmannet al., “Palm: Scal- ing language modeling with pathways,”Journal of Machine Learning Research, vol. 24, no. 240, pp. 1–113
[18]

Finetuned language models are zero-shot learners,

J. Wei, M. Bosma, V . Y . Zhao, K. Guu, A. W. Yu, B. Lester, N. Du, A. M. Dai, and Q. V . Le, “Finetuned language models are zero-shot learners,”arXiv preprint arXiv:2109.01652

arXiv
[19]

Improving cross-modal alignment for text- guided image inpainting,

Y . Zhou and G. Long, “Improving cross-modal alignment for text- guided image inpainting,” inProceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, 2023, pp. 3445–3456

2023
[20]

Visual in-context learning for large vision-language models,

Y . Zhou, X. Li, Q. Wang, and J. Shen, “Visual in-context learning for large vision-language models,” inFindings of the Association for Computational Linguistics, ACL 2024, Bangkok, Thailand and virtual meeting, August 11-16, 2024. Association for Computational Linguis- tics, ...

2024
[21]

Emobench-m: Benchmarking emotional intelligence for multimodal large language models,

H. Hu, Y . Zhou, L. You, H. Xu, Q. Wang, Z. Lian, F. R. Yu, F. Ma, and L. Cui, “Emobench-m: Benchmarking emotional intelligence for multimodal large language models,”arXiv preprint arXiv:2502.04424, 2025

2025 arXiv
[22]

Toward expert-level medical question answering with large language models,

K. Singhal, T. Tu, J. Gottweis, R. Sayres, E. Wulczyn, M. Amin, L. Hou, K. Clark, S. R. Pfohl, H. Cole-Lewiset al., “Toward expert-level medical question answering with large language models,”Nature Medicine, pp. 1–8
[23]

Capabilities of gpt-4 on medical challenge problems,

H. Nori, N. King, S. M. McKinney, D. Carignan, and E. Horvitz, “Capabilities of gpt-4 on medical challenge problems,”arXiv preprint arXiv:2303.13375

arXiv
[24]

Dr. bench: Diagnostic reasoning benchmark for clinical natural language processing,

Y . Gao, D. Dligach, T. Miller, J. Caskey, B. Sharma, M. M. Churpek, and M. Afshar, “Dr. bench: Diagnostic reasoning benchmark for clinical natural language processing,”Journal of biomedical informatics, vol. 138, p. 104286
[26]

Chatbots and large language models in radiology: a practical primer for clinical and research applications,

R. Bhayana, “Chatbots and large language models in radiology: a practical primer for clinical and research applications,”Radiology, vol. 310, no. 1, p. e232756
[27]

Large language models in simplifying radiological reports: systematic review,

Y . Artsi, V . Sorin, E. Konen, B. S. Glicksberg, G. Nadkarni, and E. Klang, “Large language models in simplifying radiological reports: systematic review,”medRxiv, pp. 2024–01

2024
[28]

Diffusion model with representation alignment for protein inverse folding,

C. Wang, Y . Zhou, Z. Zhai, J. Shen, and K. Zhang, “Diffusion model with representation alignment for protein inverse folding,”arXiv preprint arXiv:2412.09380, 2024

2024 arXiv
[29]

Adapted large language models can outperform medical experts in clinical text summarization,

D. Van Veen, C. Van Uden, L. Blankemeier, J.-B. Delbrouck, A. Aali, C. Bluethgen, A. Pareek, M. Polacin, E. P. Reis, A. Seehofnerov ´aet al., “Adapted large language models can outperform medical experts in clinical text summarization,”Nature medicine, vol. 30, no. 4, pp. 1134– 1142
[30]

Large language models in health care: Development, applications, and challenges,

R. Yang, T. F. Tan, W. Lu, A. J. Thirunavukarasu, D. S. W. Ting, and N. Liu, “Large language models in health care: Development, applications, and challenges,”Health Care Science, vol. 2, no. 4, pp. 255–263
[31]

Embracing large language models for medical applications: opportunities and challenges,

M. Karabacak and K. Margetis, “Embracing large language models for medical applications: opportunities and challenges,”Cureus, vol. 15, no. 5. 9

Pith tools

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

[1] [1]

Mimic-cxr, a de- identified publicly available database of chest radiographs with free-text reports,

A. E. Johnson, T. J. Pollard, S. J. Berkowitz, N. R. Greenbaum, M. P. Lungren, C.-y. Deng, R. G. Mark, and S. Horng, “Mimic-cxr, a de- identified publicly available database of chest radiographs with free-text reports,”Scientific data, vol. 6, no. 1, p. 317

work page

[2] [2]

Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison,

J. Irvin, P. Rajpurkar, M. Ko, Y . Yu, S. Ciurea-Ilcus, C. Chute, H. Mark- lund, B. Haghgoo, R. Ball, K. Shpanskayaet al., “Chexpert: A large chest radiograph dataset with uncertainty labels and expert comparison,” inProceedings of the AAAI conference on artificial intelligence, vol. 33, no. 01, pp. 590–597

work page

[3] [3]

Enhancing intent understanding for ambiguous prompts through human-machine co-adaptation,

Y . He, J. Wang, K. Li, Y . Wang, L. Sun, J. Yin, M. Zhang, and X. Wang, “Enhancing intent understanding for ambiguous prompts through human-machine co-adaptation,”arXiv preprint arXiv:2501.15167, 2025

arXiv 2025

[4] [4]

Natural language processing in radiology: a systematic review,

E. Pons, L. M. Braun, M. M. Hunink, and J. A. Kors, “Natural language processing in radiology: a systematic review,”Radiology, vol. 279, no. 2, pp. 329–343

work page

[5] [5]

Chexbert: combining automatic labelers and expert annota- tions for accurate radiology report labeling using bert,

A. Smit, S. Jain, P. Rajpurkar, A. Pareek, A. Y . Ng, and M. P. Lungren, “Chexbert: combining automatic labelers and expert annota- tions for accurate radiology report labeling using bert,”arXiv preprint arXiv:2004.09167

arXiv 2004

[6] [6]

A systematic review of natural language processing applied to radiology reports,

A. Casey, E. Davidson, M. Poon, H. Dong, D. Duma, A. Grivas, C. Grover, V . Su ´arez-Paniagua, R. Tobin, W. Whiteleyet al., “A systematic review of natural language processing applied to radiology reports,”BMC medical informatics and decision making, vol. 21, no. 1, p. 179

work page

[7] [7]

Utilizing bert for biomedical and clinical text mining,

R. Zhu, X. Tu, and J. X. Huang, “Utilizing bert for biomedical and clinical text mining,” inData analytics in biomedical engineering and healthcare. Elsevier, pp. 73–103

work page

[8] [8]

Thread of thought unraveling chaotic contexts,

Y . Zhou, X. Geng, T. Shen, C. Tao, G. Long, J.-G. Lou, and J. Shen, “Thread of thought unraveling chaotic contexts,”arXiv preprint arXiv:2311.08734, 2023

arXiv 2023