Pith. sign in

REVIEW 5 major objections 5 minor 38 references

Ontology-Constrained Generation of Domain-Specific Clinical Summaries

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read Ontology-guided decoding makes clinical summaries more domain-specific and less hallucinated.

desk verdict Novel ontology-guided beam scoring for clinical summarization; promising pipeline, but the headline evaluation is partly circular and needs a serious referee before strong claims. read the letter →

arxiv 2411.15666 v1 pith:KK3INE57 submitted 2024-11-23 cs.CL cs.AI

classification cs.CLcs.AI
keywords clinicalsummarizationontology-guideddecodinghallucinationreductionSNOMED-CTconstrainedgenerationdomainadaptationelectronichealthrecordsbeamsearch
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 sets out to show that a medical ontology can be used at decoding time to make LLM-generated clinical summaries more relevant to a chosen specialty and less likely to invent information. It proposes generating structured concept-level extractions from electronic health records with ontology-guided constrained decoding, then pruning those extractions to a domain and verbalizing them. On MIMIC-III the method raises the domain score for Phi-3 from 0.41 with greedy search to 0.86, and raises groundedness from 0.83 to 0.90, with similar but smaller gains for Zephyr. The hope is that clinicians can receive tailored, queryable summaries without fine-tuning a model per specialty.

What carries the argument

The engine is an ontology-guided beam search. After each generation window of ten tokens, the newly generated text is tagged with the same MedCAT annotator used on the source notes, and each beam receives a score with three components: a hierarchy score rewarding concepts that are descendants of the prompt's base ontology class, a property score rewarding mentions of classes linked to the base class by restriction properties, and a similarity score measuring ROUGE-2 overlap with the original note. The final beam score is LogSoftmax(H + P + S). A Domain-Class-Frequency dictionary, built by annotating domain texts and normalizing across domains, then drives the pruning step that selects which extracted concepts appear in the final summary.

What would settle it

Re-run the generation on the same test set with MedCAT replaced by an independent clinical concept annotator; if the domain-score and hallucination-score gaps between constrained and greedy decoding shrink to near zero, the reported improvements are an artifact of the scorer and evaluator sharing the same annotator.

Watch

Extended reading notes

Core claim

The paper's central claim is that constraining an LLM's generation to align with ontology concepts and relationships reduces the likelihood of generating information that contradicts established domain knowledge, while also making the output more relevant to a target clinical domain. The evidence comes from MIMIC-III: the ontology-constrained method more than doubles the domain score for Phi-3 compared with greedy search (0.86 versus 0.41), improves groundedness from 0.83 to 0.90, and lowers the hallucination score on the discharge-summary task. The same pipeline yields both a structured, queryable representation of each clinical note and an unstructured summary, and the pruning step lets one extraction be adapted to multiple specialties without regenerating.

Load-bearing premise

The whole pipeline and its evaluation assume that MedCAT reliably maps clinical text to SNOMED-CT concepts in both the source notes and the generated beams.

Editorial extensions

If this is right

  • Domain-specific summaries can be produced without explicitly prompting the model for a specialty; the pruning step performs the adaptation using only the ontology and the extracted concepts.
  • Hallucination scores on the Brief Hospital Course task drop for Phi-3, indicating that structured extraction plus ontology-based beam scoring is a viable faithfulness lever for smaller models.
  • The same extraction pass can feed multiple domains, since adaptation happens after extraction by pruning with the domain dictionary.
  • The method works with any LLM that exposes token probabilities, because it operates entirely at decoding time rather than requiring fine-tuning.
  • Treating the Brief Hospital Course section as a domain shows that the notion of a domain can extend beyond note categories to arbitrary target summaries.

Reading between the lines

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

  • Inference: The measured gains likely depend on how consistently MedCAT tags both the source notes and the generated beams; if the annotator is biased toward the model's preferred phrasing, the reported hallucination reduction could be inflated.
  • Inference: Because the approach only needs token probabilities and an ontology with an annotator, it could transfer to other structured knowledge domains, provided a reliable concept annotator exists.
  • Inference: Decoupling extraction from adaptation suggests a practical deployment pattern in which each admission is summarized once and then filtered per specialty, reducing redundant computation at the cost of a heavy one-time extraction.
  • Inference: The main barrier to clinical use is computational overhead from many inference passes and beam search; a testable extension would reduce the number of generation windows or cache beam scores across notes.
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. The paper proposes an ontology-guided constrained decoding method for generating domain-specific clinical summaries from electronic health records. The approach uses SNOMED-CT and the MedCAT annotator to tag clinical notes, builds a Domain-Class-Frequency (DCF) dictionary from MIMIC-III CATEGORY labels to prune extracted structured representations to a target domain, and steers beam search with hierarchy, property, and lexical-similarity scores. Experiments on MIMIC-III report large improvements in a learned domain score (e.g., Phi-3 from 0.41 to 0.86), modest improvements in NLI-based groundedness and relevance, and reductions in hallucination scores for the BHC summarization task. The authors acknowledge limitations including computational overhead, hyperparameter sensitivity, reliance on a good annotator, and the absence of human assessment.

Significance. If the reported results are trustworthy, the paper contributes a novel use of ontologies for constrained decoding that is model-agnostic and produces interpretable, structured summaries. The ability to adapt summaries to different medical specialties without prompt engineering is practically valuable. The paper is also transparent about several limitations. However, the current evidence is weakened by evaluation circularity: the domain score metric is trained on the same MIMIC-III CATEGORY labels that define the DCF pruning signal, and the hallucination metrics use the same MedCAT annotator that steers decoding. The lack of independent validation (manual or with a different annotator) makes the central claims about domain adaptation and hallucination reduction not yet firmly established.

major comments (5)
  1. [§3.1 and §5.1, Table 1] The domain score evaluation is partly circular. The DCF dictionary is constructed by annotating MIMIC-III notes with MedCAT and using the CATEGORY column to group notes by domain, while the evaluator is a BERT classifier trained on the same CATEGORY labels. The pruning step directly uses the DCF classes, so the method is rewarded for producing summaries that match the evaluator's training signal. The large improvement in Table 1 (e.g., 0.41 to 0.86 for Phi-3) may therefore be inflated. The authors should either use a human evaluation, an evaluator trained on independent labels, or an ablation that removes the overlap between the DCF construction and the evaluator training data to show that the domain score reflects genuine domain adaptation.
  2. [§3.3 and §5.3 (also §5.2)] The same MedCAT annotator is used both to compute the hierarchy and property scores that steer decoding (Section 3.3) and to compute the hallucination scores HS and AHS (Section 5.3) and the groundedness evaluation (Section 5.2). If MedCAT is noisy, biased toward certain phrasings, or imperfect in the target domains, the reported hallucination reduction (e.g., HS from 45.70 to 37.95 for Phi-3 in Table 3) could reflect annotator alignment rather than true clinical faithfulness. The authors' own concession in Section 6.4 that automated evaluation 'may not be always trustworthy' applies directly to these metrics. An independent annotation study, ideally with clinician involvement or at least a second, different concept annotator, is needed to validate the hallucination-reduction claim.
  3. [§3.3, Table 2] The contribution of the ontology-based scores is not isolated. The similarity score S (ROUGE-2 to the clinical note, with Sbf = 10) alone could improve groundedness by favoring beams that lexically resemble the note. Without an ablation that varies the presence of H, P, and S, it is unclear whether the observed groundedness improvements in Table 2 are due to the ontology constraints or simply to the lexical-similarity term. The authors should report results with S only, H+P only, and the full combination.
  4. [§5.2, Table 2] The groundedness evaluation relies on a DEBERTA model fine-tuned on MNLI, which is not trained on medical data. The authors note this in Section 6.4, but the claim in Section 6.2 that 'summaries generated using our custom ontology-guided decoding process are more likely to be factually consistent' is not supported by a medically validated NLI model. Additionally, no confidence intervals or significance tests are reported; with 1000 samples, the difference between 0.83 and 0.90 for Phi-3 groundedness may or may not be statistically reliable. Please report error bars or a significance test.
  5. [§5.3, Eqs. (2)-(3), Table 3] The hallucination scores HS and AHS depend on the sets of concepts extracted by MedCAT from the notes and the generated summary. Beyond the circularity concern, this metric can be sensitive to annotator precision and recall in both the source and the summary. The adjusted score AHS uses the reference summary to allow inferred concepts, but it still relies on the same annotator. The authors should provide evidence that MedCAT's annotations are reliable in this setting, or supplement the automated scores with human evaluation, before claiming that the method reduces hallucinations.
minor comments (5)
  1. [§6.3 and §6.4] There are typos: 'is is not surprising' in Section 6.3 and 'highly sensible to hyperparameters' in Section 6.4 should be 'highly sensitive'.
  2. [§5.3, Eqs. (2)-(3)] The notation is confusing: N is used for the number of samples in Eq. (1) and for the set of concepts in the notes in Eqs. (2)-(3). Please use distinct symbols to avoid ambiguity.
  3. [Figures 1 and 2] The figure captions are incomplete and do not explain all the abbreviations and arrows (e.g., what 'CSR' stands for in Figure 3 is defined, but the captions for Figures 1 and 2 could be more explicit).
  4. [References] Some references contain malformed publisher fields, e.g., references [2] and [16] include '[object Object]' in the publisher entry. These should be cleaned up.
  5. [§1] The claim that this is the 'first approach that employs ontologies in conjunction with LLMs to constrain generation' is strong. Given existing work on knowledge-graph-constrained decoding (e.g., reference [6]), it would be safer to phrase this as 'to the best of our knowledge' and to clarify the distinction between ontologies and knowledge graphs.

Circularity Check

2 steps flagged · score 4.0 of 10

Evaluation partly circular: the domain score uses the same MIMIC-III CATEGORY labels that build the DCF, and HS/AHS shares the MedCAT annotator with the constrained decoder; NLI groundedness provides independent partial support.

  1. fitted input called prediction [Section 5.1 (Domain adaptation evaluator, Eq. 1) vs. Sections 3.1 and 3.4 (DCF construction and pruning)]
    "To do so, we utilized the CATEGORY column of MIMIC-III which indicates the medical field of a clinical note. In practice, since not all domains are equally present, we focus on the Nursing, ECG, Radiology and Physician domains as they are the most frequent."

    The DCF that drives pruning is built from texts 'linked to a certain domain D' using the same MIMIC-III CATEGORY-derived domains, and the pruning step keeps only the top-k classes of that domain's DCF. The evaluator used for the domain score D is then fine-tuned on those same CATEGORY labels. The result in Table 1 (0.86 vs 0.41) therefore measures how well a summary pruned to domain D matches a classifier trained to recognize domain D; it is not an independent validation that the summaries are clinically more relevant to the specialty. The prediction target is aligned with the construction target by design.

  2. other [Section 4.2 (annotator) and Section 5.3, Eqs. (2)-(3), with Section 3.3 constrained decoding]
    "As for the annotator linking text sequences and SNOMED-CT classes, we utilized the MedCAT annotator [15]. This annotator was used during the decoding process and during evaluation to tag concepts (see Section 5.3)."

    The constrained decoder scores beam candidates using MedCAT-tagged classes (hierarchy and property scores, Section 3.3), and the hallucination scores HS and AHS are computed from MedCAT concept sets N and S (Section 5.3). The same annotator is thus both a component of the generation mechanism and the measurement instrument for the hallucination claim. Any systematic tagging bias is shared by the control and the metric, so the HS/AHS improvements in Table 3 are not an independent check on factual grounding. The authors acknowledge the dependency in Section 6.4: the approach 'heavily relies on a good annotator' and automated evaluation 'may not be always trustworthy.'

full rationale

The paper's derivation chain is not fully circular: the constrained decoding procedure is well-defined, and Table 2's groundedness scores use a separate DEBERTA NLI model, giving independent (though admittedly not medically fine-tuned) evidence that the ontology-guided decoding improves faithfulness. There are no load-bearing self-citations or imported uniqueness theorems. The partial circularity is confined to two evaluation components. First, the domain-adaptation score is partly self-consistent: the target category used in Eq. (1) is the same MIMIC-III CATEGORY label that determines the DCF used for pruning, so the large jump in Table 1 is expected by construction. Second, the HS/AHS hallucination metrics share the MedCAT annotator with the beam-scoring function, so they do not independently confirm the hallucination-reduction effect. These issues are explicitly flagged by the authors in Section 6.4 as reliance on the annotator and lack of human/gold-standard evaluation. Because the central hallucination claim retains independent NLI support and no self-citation chain is load-bearing, a moderate score of 4 is appropriate rather than a higher score.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The approach is built on existing resources (SNOMED-CT, MedCAT, MIMIC-III, pretrained LLMs) plus eight fitted or hand-chosen hyperparameters. The most load-bearing assumptions are that MedCAT and the automated evaluators measure what they claim to measure, which the authors themselves question in Section 6.4. No new physical or conceptual entities are postulated; the DCF dictionary is a data structure, not an entity.

free parameters (8)
  • Hierarchy boost factor (Hbf) = 3
    Controls how strongly beams are rewarded for containing descendant concepts of the base class; chosen empirically.
  • Property boost factor (Pbf) = 10
    Controls reward for mentioning classes linked via restriction properties; chosen empirically.
  • Similarity boost factor (Sbf) = 10
    Controls reward for ROUGE-2 similarity between beam and clinical note; chosen empirically.
  • Beam size and group size = 10 and 2
    Decoding hyperparameters for diverse beam search.
  • Generation window = 10 tokens
    Number of tokens before beam scoring is applied; authors state 5 to 15 is a good range and they use 10.
  • Pruning top-k classes = 30
    Controls how many frequent DCF classes are kept in the structured summary.
  • Neighbor depth alpha = 2
    Controls how many nearby ontology subclasses are added around frequent classes during pruning.
  • DCF minimum occurrence threshold = not specified
    Section 3.1 builds S based on a minimum occurrence threshold but no value is given.
assumptions (5)
  • domain assumption SNOMED-CT ontology correctly represents the medical knowledge needed for the target domains.
    The whole pipeline uses SNOMED-CT hierarchy and restriction properties to score beams and prune classes; incorrect or incomplete ontology coverage would misguide generations.
  • domain assumption MedCAT annotator correctly detects SNOMED-CT concepts and their surface variants in clinical text.
    The annotator is used in the extraction, in the constrained decoding scores, and in the hallucination metrics; the authors note the method depends on a good annotator.
  • domain assumption ROUGE-2 is an adequate similarity measure between a beam's generated text and the clinical note for grounding purposes.
    The similarity score S = Sbf * ROUGE-2 assumes textual overlap is a good proxy for factual grounding during decoding.
  • domain assumption DEBERTA-MNLI entailment scores are valid measures of groundedness and relevance for clinical summaries.
    The hallucination evaluation relies on this NLI model; the authors concede it was not trained on medical data and may not be trustworthy.
  • domain assumption The BERT classifier's predicted domain probability is a valid measure of domain adaptation.
    Domain score D averages the evaluator logit for the expected domain; the evaluator was trained on the same CATEGORY labels used to build the DCF, making the metric partly circular.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Ontology-Constrained Generation of Domain-Specific Clinical Summaries." pith.science (2026). https://pith.science/paper/KK3INE57

@misc{pith2026241115666,
  author       = {Pith},
  title        = {Pith review of: Ontology-Constrained Generation of Domain-Specific Clinical Summaries},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KK3INE57}},
  note         = {Machine review of arXiv:2411.15666}
}
read the original abstract

Large Language Models (LLMs) offer promising solutions for text summarization. However, some domains require specific information to be available in the summaries. Generating these domain-adapted summaries is still an open challenge. Similarly, hallucinations in generated content is a major drawback of current approaches, preventing their deployment. This study proposes a novel approach that leverages ontologies to create domain-adapted summaries both structured and unstructured. We employ an ontology-guided constrained decoding process to reduce hallucinations while improving relevance. When applied to the medical domain, our method shows potential in summarizing Electronic Health Records (EHRs) across different specialties, allowing doctors to focus on the most relevant information to their domain. Evaluation on the MIMIC-III dataset demonstrates improvements in generating domain-adapted summaries of clinical notes and hallucination reduction.

Figures

Figures reproduced from arXiv: 2411.15666 by the authors.

Figure 1
Figure 1. Overview of our general architecture to generate domain-tailored summaries. 2 Related work Summarization. Summarization is the process of generating a smaller text from a larger input text. The main objective of that process is to grasp the information of the input 3 Code is available at https://github.com/Lama-West/Ontology-based-decoding_EKAW2024 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall architecture of our method : Multiple notes about the same patient are passed to the framework and structured and unstructured summaries are generated 3.1 Domain adaptation analysis We define a domain to be a set of ontology classes of interest related to a specific med￾ical field. To adapt the generation to multiple domains, an initial analysis is performed [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Extraction Phase To improve the information extraction capabilities of the LLM, we adapted [32]’s summarization technique to the medical domain by incorporating an ontology-guided prompting process. Given multiple clinical notes of a patient during an admission and a medical ontology, we start by annotating each clinical note to retrieve all the medical [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Constrained decoding process : Each beam rectangle represents the current generation window associated with the beam. The concepts in green are concepts that are associated to a children class of the base class (Drug or medicament) in the ontology. Green concepts impro…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 10 canonical work pages

  1. [1]

    Abdin, M., Jacobs, S.A., et al.: Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone (May 2024), http://arxiv.org/abs/2404.14219, arXiv:2404.14219 [cs]

  2. [2]

    https://doi.org/10.48550/ ARXIV.2401.02369, https://arxiv.org/abs/2401.02369, publisher: [object Object] Version Number: 1

    Adams, G., Zucker, J., Elhadad, N.: SPEER: Sentence-Level Planning of Long Clinical Summaries via Embedded Entity Retrieval (2024). https://doi.org/10.48550/ ARXIV.2401.02369, https://arxiv.org/abs/2401.02369, publisher: [object Object] Version Number: 1

  3. [3]

    03948, arXiv:2303.03948 [cs]

    Adams, G., Zucker, J., Elhadad, N.: A Meta-Evaluation of Faithfulness Metrics for Long- Form Hospital-Course Summarization (Mar 2023), http://arxiv.org/abs/2303. 03948, arXiv:2303.03948 [cs]

  4. [4]

    org/abs/1904.03323, arXiv:1904.03323 [cs]

    Alsentzer, E., Murphy, J.R., Boag, W., Weng, W.H., Jin, D., Naumann, T., McDermott, M.B.A.: Publicly Available Clinical BERT Embeddings (Jun 2019), http://arxiv. org/abs/1904.03323, arXiv:1904.03323 [cs]

  5. [5]

    org/abs/2307.04507, arXiv:2307.04507 [cs]

    Chern, I.C., Wang, Z., Das, S., Sharma, B., Liu, P., Neubig, G.: Improving Factuality of Abstractive Summarization via Contrastive Reward Learning (Jul 2023),http://arxiv. org/abs/2307.04507, arXiv:2307.04507 [cs]

  6. [6]

    Choi, S., Fang, T., Wang, Z., Song, Y .: KCTS: Knowledge-Constrained Tree Search Decod- ing with Token-Level Hallucination Detection (Oct 2023), http://arxiv.org/abs/ 2310.09044, arXiv:2310.09044 [cs]

  7. [7]

    Chuang, Y .N., Tang, R., Jiang, X., Hu, X.: SPeC: A Soft Prompt-Based Calibration on Perfor- mance Variability of Large Language Model in Clinical Notes Summarization (Aug 2023), http://arxiv.org/abs/2303.13035, arXiv:2303.13035 [cs]

  8. [8]

    Cohen, R., Hamri, M., Geva, M., Globerson, A.: LM vs LM: Detecting Factual Er- rors via Cross Examination (May 2023), http://arxiv.org/abs/2305.13281, arXiv:2305.13281 [cs]

Show all 38 references
  1. [9]

    11495, publisher: arXiv Version Number: 2

    Dhuliawala, S., Komeili, M., Xu, J., Raileanu, R., Li, X., Celikyilmaz, A., Weston, J.: Chain-of-Verification Reduces Hallucination in Large Language Models (2023).https:// doi.org/10.48550/ARXIV.2309.11495, https://arxiv.org/abs/2309. 11495, publisher: arXiv Version Number: 2

  2. [10]

    Gao, Y ., Li, R., Caskey, J., Dligach, D., Miller, T., Churpek, M.M., Afshar, M.: Leveraging A Medical Knowledge Graph into Large Language Models for Diagnosis Prediction (Aug 2023), http://arxiv.org/abs/2308.14321, arXiv:2308.14321 [cs]

  3. [11]

    13971, arXiv:2305.13971 [cs]

    Geng, S., Josifoski, M., Peyrard, M., West, R.: Grammar-Constrained Decoding for Struc- tured NLP Tasks without Finetuning (Nov 2023), http://arxiv.org/abs/2305. 13971, arXiv:2305.13971 [cs]

  4. [12]

    Guerreiro, N.M., Alves, D., Waldendorf, J., Haddow, B., Birch, A., Colombo, P., Mar- tins, A.F.T.: Hallucinations in Large Multilingual Translation Models (Mar 2023), http: //arxiv.org/abs/2303.16104, arXiv:2303.16104 [cs]

  5. [13]

    In: International Conference on Learning Representations (2021), https:// openreview.net/forum?id=XPZIaotutsD

    He, P., Liu, X., Gao, J., Chen, W.: Deberta: Decoding-enhanced bert with disentangled attention. In: International Conference on Learning Representations (2021), https:// openreview.net/forum?id=XPZIaotutsD

  6. [14]

    Scientific Data3(1), 160035 (May 2016).https://doi.org/10.1038/ sdata.2016.35, https://www.nature.com/articles/sdata201635

    Johnson, A.E., Pollard, T.J., Shen, L., Lehman, L.w.H., Feng, M., Ghassemi, M., Moody, B., Szolovits, P., Anthony Celi, L., Mark, R.G.: MIMIC-III, a freely accessible critical care database. Scientific Data3(1), 160035 (May 2016).https://doi.org/10.1038/ sdata.2016.35, https:/...

  7. [15]

    Kraljevic, Z., Searle, T., Shek, A., Roguski, L., Noor, K., Bean, D., Mascio, A., Zhu, L., Folarin, A.A., Roberts, A., Bendayan, R., Richardson, M.P., Stewart, R., Shah, A.D., Wong, Ontology-Constrained Generation of Domain-Specific Clinical Summaries 17 W.K., Ibrahim, Z., Teo...

  8. [16]

    https://doi.org/10.48550/ARXIV

    Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V ., Goyal, N., Küttler, H., Lewis, M., Yih, W.t., Rocktäschel, T., Riedel, S., Kiela, D.: Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (2020). https://doi.org/10.48550/ARXIV. 2005.11401, https://arx...

  9. [17]

    Li, K., Patel, O., Viégas, F., Pfister, H., Wattenberg, M.: Inference-Time Intervention: Elic- iting Truthful Answers from a Language Model (Oct 2023), http://arxiv.org/abs/ 2306.03341, arXiv:2306.03341 [cs]

  10. [18]

    In: The 22nd Workshop on Biomedical Natu- ral Language Processing and BioNLP Shared Tasks

    Liu, M., Zhang, D., Tan, W., Zhang, H.: DeakinNLP at ProbSum 2023: Clinical Progress Note Summarization with Rules and Language ModelsClinical Progress Note Summariza- tion with Rules and Languague Models. In: The 22nd Workshop on Biomedical Natu- ral Language Processing and B...

  11. [19]

    org/abs/1906.07241, arXiv:1906.07241 [cs]

    Logan IV , R.L., Liu, N.F., Peters, M.E., Gardner, M., Singh, S.: Barack’s Wife Hillary: Us- ing Knowledge-Graphs for Fact-Aware Language Modeling (Jun 2019),http://arxiv. org/abs/1906.07241, arXiv:1906.07241 [cs]

  12. [20]

    Lu, X., West, P., Zellers, R., Bras, R.L., Bhagavatula, C., Choi, Y .: NeuroLogic Decod- ing: (Un)supervised Neural Text Generation with Predicate Logic Constraints (Apr 2021), http://arxiv.org/abs/2010.12884, arXiv:2010.12884 [cs]

  13. [21]

    Press, O., Zhang, M., Min, S., Schmidt, L., Smith, N.A., Lewis, M.: Measuring and Narrow- ing the Compositionality Gap in Language Models (Oct 2023), http://arxiv.org/ abs/2210.03350, arXiv:2210.03350 [cs]

  14. [22]

    Journal of Biomedical Informatics 141, 104358 (May 2023)

    Searle, T., Ibrahim, Z., Teo, J., Dobson, R.J.: Discharge summary hospital course summari- sation of in patient Electronic Health Record text with clinical concept guided deep pre- trained Transformer models. Journal of Biomedical Informatics 141, 104358 (May 2023). https://do...

  15. [23]

    Ar- tificial Intelligence in Medicine 123, 102224 (Jan 2022)

    Shi, J., Gao, X., Kinsman, W.C., Ha, C., Gao, G.G., Chen, Y .: DI++: A deep learning system for patient condition identification in clinical notes. Ar- tificial Intelligence in Medicine 123, 102224 (Jan 2022). https://doi.org/ 10.1016/j.artmed.2021.102224, https://linkinghub.e...

  16. [24]

    Physicians

    Sorita, A., Robelia, P.M., Kattel, S.B., McCoy, C.P., Keller, A.S., Almasri, J., Mu- rad, M.H., Newman, J.S., Kashiwagi, D.T.: The Ideal Hospital Discharge Sum- mary: A Survey of U.S. Physicians. Journal of Patient Safety 17(7) (2021), https: //journals.lww.com/journalpatients...

  17. [25]

    Proc AMIA Symp pp

    Stearns, M.Q., Price, C., Spackman, K.A., Wang, A.Y .: SNOMED clinical terms: overview of the development process and project status. Proc AMIA Symp pp. 662–666 (2001)

  18. [26]

    Stengel-Eskin, E., Rawlins, K., Van Durme, B.: Zero and Few-shot Semantic Pars- ing with Ambiguous Inputs (Jan 2024), http://arxiv.org/abs/2306.00824, arXiv:2306.00824 [cs]

  19. [27]

    J Med Internet Res 22(7), e19274 (Jul 2020)

    Tajirian, T., Stergiopoulos, V ., Strudwick, G., Sequeira, L., Sanches, M., Kemp, J., Ra- mamoorthi, K., Zhang, T., Jankowicz, D.: The influence of electronic health record use on physician burnout: Cross-sectional survey. J Med Internet Res 22(7), e19274 (Jul 2020). https://d...

  20. [28]

    Tunstall, L., Beeching, E., Lambert, N., Rajani, N., Rasul, K., Belkada, Y ., Huang, S., von Werra, L., Fourrier, C., Habib, N., Sarrazin, N., Sanseviero, O., Rush, A.M., Wolf, T.: Zephyr: Direct Distillation of LM Alignment (Oct 2023), http://arxiv.org/abs/ 2310.16944, arXiv:...

  21. [30]

    Van Veen, D., Van Uden, C., Blankemeier, L., Delbrouck, J.B., Aali, A., Bluethgen, C., Pareek, A., Polacin, M., Reis, E.P., Seehofnerova, A., Rohatgi, N., Hosamani, P., Collins, W., Ahuja, N., Langlotz, C.P., Hom, J., Gatidis, S., Pauly, J., Chaudhari, A.S.: Adapted Large Lang...

  22. [31]

    Vijayakumar, A.K., Cogswell, M., Selvaraju, R.R., Sun, Q., Lee, S., Crandall, D., Batra, D.: Diverse Beam Search: Decoding Diverse Solutions from Neural Sequence Models (Oct 2018), http://arxiv.org/abs/1610.02424, arXiv:1610.02424 [cs]

  23. [32]

    Wang, Y ., Zhang, Z., Wang, R.: Element-aware Summarization with Large Language Models: Expert-aligned Evaluation and Chain-of-Thought Method (May 2023), http: //arxiv.org/abs/2305.13412, arXiv:2305.13412 [cs]

  24. [33]

    https://doi.org/10.48550/ARXIV.2201.11903, https://arxiv

    Wei, J., Wang, X., Schuurmans, D., Bosma, M., Ichter, B., Xia, F., Chi, E., Le, Q., Zhou, D.: Chain-of-Thought Prompting Elicits Reasoning in Large Language Models (2022). https://doi.org/10.48550/ARXIV.2201.11903, https://arxiv. org/abs/2201.11903, publisher: [object Object] ...

  25. [34]

    Willard, B.T., Louf, R.: Efficient Guided Generation for Large Language Models (Aug 2023), http://arxiv.org/abs/2307.09702, arXiv:2307.09702 [cs]

  26. [35]

    In: Proceedings of the 2018 Conference of the North Ameri- can Chapter of the Association for Computational Linguistics: Human Language Technolo- gies, V olume 1 (Long Papers)

    Williams, A., Nangia, N., Bowman, S.: A broad-coverage challenge corpus for sentence un- derstanding through inference. In: Proceedings of the 2018 Conference of the North Ameri- can Chapter of the Association for Computational Linguistics: Human Language Technolo- gies, V olu...

  27. [36]

    Wu, Z., Hu, Y ., Shi, W., Dziri, N., Suhr, A., Ammanabrolu, P., Smith, N.A., Ostendorf, M., Hajishirzi, H.: Fine-Grained Human Feedback Gives Better Rewards for Language Model Training (Oct 2023), http://arxiv.org/abs/2306.01693, arXiv:2306.01693 [cs]

  28. [37]

    In: Proceedings of the 2021 Conference of the North American Chapter of the Associa- tion for Computational Linguistics: Human Language Technologies

    Yang, K., Klein, D.: FUDGE: Controlled Text Generation With Future Discriminators. In: Proceedings of the 2021 Conference of the North American Chapter of the Associa- tion for Computational Linguistics: Human Language Technologies. pp. 3511–3535 (2021). https://doi.org/10.186...

  29. [38]

    van Zandvoort, D., Wiersema, L., Huibers, T., van Dulmen, S., Brinkkemper, S.: En- hancing Summarization Performance through Transformer-Based Prompt Engineering in Automated Medical Reporting (Jan 2024), http://arxiv.org/abs/2311.13274, arXiv:2311.13274 [cs]

  30. [39]

    Zheng, L., Yin, L., Xie, Z., Sun, C., Huang, J., Yu, C.H., Cao, S., Kozyrakis, C., Sto- ica, I., Gonzalez, J.E., Barrett, C., Sheng, Y .: SGLang: Efficient Execution of Struc- tured Language Model Programs (Jun 2024),http://arxiv.org/abs/2312.07104, arXiv:2312.07104 [cs]

Pith tools

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