Pith. sign in

REVIEW 3 major objections 4 minor 30 references

Endoscopic VQA models can score high on complex answers while failing the component questions on the same image, and a training-free method that reconciles the two narrows the gap.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 16:50 UTC pith:AN3LDZKG

load-bearing objection The inconsistency finding is real and the ASR ablation is honest; EndoCA is a useful benchmark for endoscopic VQA, but it needs a public release and human validation of its label-conformance scorer before the reliability claims fully convince. the 3 major comments →

arxiv 2607.17834 v1 pith:AN3LDZKG submitted 2026-07-20 cs.CV cs.AI

Measuring and Improving Complex-Atomic Answer Consistency in Endoscopic VQA

Yuhao Liu , Cheng Zhao , Guanghui Yue This is my paper
classification cs.CV cs.AI
keywords endoscopic VQAcomplex-atomic consistencyanswer consistencyvision-language modelsjoint accuracyselective answeringmedical visual question answeringanswer revision
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper's central claim is that in endoscopic visual question answering, a model's accuracy on complex multi-part questions can be high while its answers to the component atomic questions on the same image are often wrong, so complex-answer scores overstate reliability. To make this visible, the authors build a paired benchmark in which each complex question is evaluated together with its associated same-image atomic questions, and report that one strong endoscopy-adapted model reaches 74.5% complex accuracy but only 55.8% joint accuracy, with a quarter of its accepted complex answers contradicting its own atomic predictions. They then propose a training-free method, Atomic-Support Reconciliation, that feeds the model's own atomic answers back as premises for revising the complex answer or for abstaining. Across four models it raises joint accuracy by 3.1 to 8.1 percentage points and cuts complex-atomic inconsistency by an average of 10.4 points. A sympathetic reader would care because the result implies that current final-answer benchmarks in this medical imaging domain can hide exactly the kind of unstable evidence-grounded reasoning that matters clinically.

Core claim

The paper introduces EndoCA, a paired complex-atomic consistency benchmark built from an existing two-level endoscopic VQA resource, and uses it to argue that complex-answer accuracy alone is an incomplete signal. On 11 vision-language models, the benchmark's joint accuracy—requiring the complex answer and every separately generated atomic answer to be correct—runs well below complex-answer accuracy, and its complex-atomic inconsistency metric shows that a substantial share of accepted complex answers are accompanied by at least one wrong atomic prediction on the same image. The same paired signal motivates ASR, a training-free reconciliation mechanism: the model first answers the atomic que

What carries the argument

The paired evaluation unit—an image, a complex question, its complex answer annotation, and an associated list of atomic question-answer pairs—with a fixed question-type-aware scorer and four metrics: complex-answer accuracy, atomic-answer accuracy, joint accuracy, and complex-atomic inconsistency. Joint accuracy and complex-atomic inconsistency are the load-bearing quantities: they convert the question 'is the final answer right' into 'is every component judgment on the same image right, and do the two agree.' The reconciliation mechanism ASR reuses this structure, treating model-generated atomic answers as premises for a revision prompt and for a consistency-judged answer/abstain decision.

Load-bearing premise

The whole inconsistency metric assumes that every correct complex answer can be exactly decomposed into the atomic labels inherited from the source dataset, so a semantically correct paraphrase that does not align with those labels is counted as inconsistent.

What would settle it

Collect or construct a set of complex answers that clinicians judge clinically correct but that do not exactly match the inherited atomic label sets; recomputing the metrics with a lenient semantic-match rule would show whether the benchmark's gap and ASR's 10.4-point inconsistency reduction are genuine or partly artifacts of the exact-match scorer.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Endoscopic VQA evaluation should report joint accuracy and complex-atomic inconsistency alongside complex-answer accuracy whenever paired atomic questions exist; otherwise rankings and reported reliability can mislead.
  • Models with nearly identical complex-answer accuracy can differ by more than 14 percentage points in joint accuracy, so consistency-aware scores reorder model comparisons.
  • As question complexity rises—more atomic components per complex question—all-atomic accuracy falls faster than complex accuracy, making multi-part questions a stress test of paired correctness.
  • Model-generated atomic answers can serve as inference-time premises for revision without retraining, improving paired correctness on top models.
  • Consistency-guided abstention provides a reliability–coverage trade-off, letting models answer more often when atomic premises agree and abstain when they conflict.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The benchmark's inconsistency score is defined under exact normalized matching to inherited atomic labels; a lenient semantic-match scorer might shrink the reported gap, so the gap's magnitude is partly a function of scoring strictness, not only model behavior.
  • The ASR pattern—answer components, then reconcile or abstain—could transfer to other multi-component medical VQA settings such as radiology or pathology wherever atomic subquestions exist; the paper does not test this.
  • The abstention signal could double as a human-review trigger in clinical deployment: cases where atomic premises conflict are exactly the ones a clinician should re-check.
  • Because ASR uses only the model's own outputs, its gains could in principle generalize across architectures; a natural extension is to test whether fine-tuning on consistency pairs shrinks the gap more durably than inference-time reconciliation.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper introduces EndoCA, a paired complex-atomic answer consistency benchmark for endoscopic VQA, built from the two-level QA structure of Kvasir-VQA and Kvasir-VQA-x1. Each test sample pairs a complex question with its associated atomic questions; complex-answer correctness is defined by the EndoCA scorer as normalized exact agreement with all associated atomic annotations. The paper evaluates 11 VLMs and reports complex accuracy, atomic accuracy, joint accuracy, and complex-atomic inconsistency. It finds that several models achieve high complex-answer accuracy while showing materially lower joint accuracy and non-negligible complex-atomic inconsistency. It then proposes Atomic-Support Reconciliation (ASR), a training-free method that uses the model's own atomic answers as contextual premises for revising complex answers and for consistency-guided selective answering. On four models, ASR-Revise improves joint accuracy by +3.1 to +8.1 percentage points and reduces complex-atomic inconsistency by 10.4 points on average; ASR-Selective trades coverage for answered-subset accuracy.

Significance. If the EndoCA scorer is accepted as valid, this is a useful benchmark contribution: it provides a reproducible, fixed-protocol metric for a clinically relevant reliability gap, evaluates a broad and representative set of VLMs, includes a controlled complexity axis, and proposes a training-free mitigation with a clean ablation that separates atomic-premise revision from generic self-checking. The paper is careful to report four complementary metrics and to distinguish complex-answer accuracy from paired correctness. However, the central claim is contingent on the validity of the EndoCA scoring rule, and the current manuscript provides no human-agreement or paraphrase-sensitivity validation for arbitrary model outputs. Since the ASR method is also evaluated with the same scoring rule, the headline 'hidden inconsistency' numbers and the ASR gains could partly reflect label-conformance rather than genuine unreliability or improvement. These concerns are addressable with targeted experiments, and the strengths of the benchmark design justify a revision rather than rejection.

major comments (3)
  1. [Section III-C, Eq. (2)-(3), Table II] Section III-C and Eq. (2)-(3) define complex-answer correctness as normalized exact agreement with every associated atomic annotation, but Table II's worked example describes the scorer as 'extract[ing] component labels' from a free-form sentence ('visible text is present' becomes 'yes'). These two descriptions conflict. If the scorer performs semantic component extraction, the exact-match wording and the claim that no additional LLM judge is used are incomplete; if it only applies surface normalization, the example overstates what is checked. This matters because the paper's headline phenomenon—complex accuracy hiding inconsistency—and the ASR gains in Table V are both measured with this scorer. No human-agreement study is reported; the source-dataset construction-time agreement does not validate scoring of arbitrary model outputs. I request a precise specification of the matching/extra
  2. [Section IV-B, Tables V-VI] ASR-Revise conditions the revision on Pi = {(qij, a_hat_ij)}, the model's own atomic answers, while the EndoCA scorer checks complex answers against the associated atomic annotations. The reported +3.1 to +8.1 joint-accuracy gains and −10.4 average inconsistency reduction may therefore reflect the revised answer conforming to the premise-label format rather than a genuine improvement in evidence use. The no-atomic ablation (Table VI) isolates the presence of atomic premises, but it does not test whether the benefit comes from independent evidence reconciliation or label-copying. Please add controls in which the atomic premises are (i) another model's atomic answers and (ii) deliberately corrupted atomic labels; report whether the revised answers follow the premise labels, and whether the joint-accuracy gain persists under human or semantic scoring.
  3. [Section V-C, Table V] ASR-Selective is evaluated only by coverage and accuracy on the answered subset, without a comparison at matched coverage. Direct has 100% coverage, so the 71.0% versus 63.3% comparison does not separate the benefit of abstaining from the benefit of selecting easy cases. The claim that ASR-Selective 'improves accuracy on answered cases' should be supported by a risk-coverage curve, by direct accuracy at the same coverage (for example, random or confidence-based selection), or by a calibrated selective-prediction metric. Without this, the selective-answering contribution is not yet fully established.
minor comments (4)
  1. [Throughout] There are LaTeX rendering issues: 'atomicquestion' should be 'atomic question', and 'LLaV A-Med' should be 'LLaVA-Med'. Please check all model names for consistent typography.
  2. [Tables III-V] No confidence intervals or significance tests are reported. Given deterministic decoding and large fixed sample sizes this may be acceptable, but the authors should state explicitly that all metrics are point estimates over a fixed benchmark and that no repeated trials were run.
  3. [Table V] Table V reports only deltas for Joint Acc. and Complex-Atomic Inconsistency. It would help the reader to also show the absolute Direct and Revised values for these two metrics, or to include a note that the Direct values are those in Table III.
  4. [Section VI] The conclusion concedes that the current protocol is tied to atomic decomposability and defers free-form clinical questions to future work. This limitation should appear earlier, in the benchmark construction or scoring section, so readers interpret the metric accordingly.

Circularity Check

0 steps flagged

No circularity: EndoCA's metrics are transparent definitions and ASR's gains are empirical; scorer validation is a correctness risk, not a circular step.

full rationale

EndoCA is a measurement protocol, not a derivation: its central quantities (C_i, A_ij, joint accuracy, inconsistency) are defined transparently in Eqs. (2)-(6). Complex-answer correctness is explicitly a label-conformance rule (Section III-C: 'A complex prediction is counted as correct only when the normalized answer matches every associated atomic answer annotation'), but the paper does not claim to derive complex correctness from atomic correctness; it empirically probes the same models with separately generated atomic prompts. The headline claim that complex-answer accuracy can hide inconsistency is a contingent empirical result from Tables III-IV, not a consequence of the definitions: nothing in Eq. (3) forces the separate atomic predictions A_ij to fail when C_i=1, and the measured inconsistencies vary widely across models (12.5%-83.1%). ASR is a post-processing method that intentionally conditions revision on model-generated atomic premises; its joint-accuracy gains are measured rather than entailed, and the ablation (Table VI) shows that removing atomic premises eliminates the gains. The lack of a human-agreement study for the scorer is a validity/correctness limitation, and the conclusion explicitly defers free-form clinical questions, but these do not make the empirical claims circular. No fitted parameter is renamed as a prediction, and no load-bearing self-citation or imported uniqueness theorem is present.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

The central claim rests on trusting the Kvasir-VQA/Kvasir-VQA-x1 source annotations and on defining complex-answer correctness as exact agreement with atomic answer annotations. No model parameters are fitted, but the benchmark sample sizes and question-complexity strata are hand-chosen, and the scoring oracle for complex answers is the same atomic-label structure that ASR uses as revision premises.

axioms (4)
  • domain assumption Kvasir-VQA and Kvasir-VQA-x1 annotations are correct ground truth for the endoscopic images, and the recorded complex-to-atomic associations are valid.
    Section III-A states EndoCA uses the recorded association linking each complex question to its atomic QA annotations; all EndoCA metrics inherit any noise or misassociation from these source datasets, and no clinician re-verification of the associations is reported.
  • ad hoc to paper A correct complex answer is fully captured by the set of associated atomic answer annotations under the fixed scorer.
    Section III-C: 'A complex prediction is counted as correct only when the normalized answer matches every associated atomic answer annotation.' This constructs the complex-answer ground truth from atomic labels and is not independently clinically validated.
  • domain assumption The fixed question-type-aware scorer's normalization rules preserve semantic equivalence for all model outputs.
    Section III-C describes normalization of case, spacing, punctuation, yes/no forms, count words, and canonical labels, but no inter-annotator agreement or human validation of the scorer's semantic equivalence is reported.
  • domain assumption Model-generated atomic answers, when provided as revision premises, do not leak the scoring structure in a way that trivially inflates ASR gains.
    ASR conditions revision on the model's own atomic answers, while the complex score is computed against the same atomic annotations; the paper does not analyze whether this coupling inflates the joint-accuracy improvement beyond genuine image understanding.

pith-pipeline@v1.3.0-alltime-deepseek · 10416 in / 13373 out tokens · 129291 ms · 2026-08-01T16:50:47.513910+00:00 · methodology

0 comments
read the original abstract

Endoscopic visual question answering (VQA) increasingly asks complex questions that combine several endoscopic answer components rather than isolated factual queries. Such complex answers may be scored as correct even when the same model fails on associated atomic questions. We introduce EndoCA, a paired complex-atomic answer consistency benchmark for evaluating whether complex answers remain consistent with same-image atomic answers. EndoCA contains two suites: EndoCA-Core evaluates compact question-complexity patterns commonly seen in practical endoscopic VQA, and EndoCA-Diagnostic supports controlled analysis across increasing question complexity. We evaluate 11 VLMs spanning open, medical, endoscopy-adapted, and closed-source models on EndoCA. Some VLMs achieve high complex-answer accuracy, yet their atomic-answer accuracy and complex-atomic answer consistency remain substantially lower. To reduce this complex-atomic inconsistency, we introduce Atomic-Support Reconciliation (ASR), a training-free mechanism that uses model-generated atomic answers as contextual premises for answer revision and consistency-guided selective answering. On four selected publicly available models, ASR-Revise improves paired complex-atomic correctness with modest changes in complex-answer accuracy, while ASR-Selective improves accuracy on answered cases by allowing the model to abstain from less reliable cases. Together, EndoCA and ASR provide a consistency-aware benchmark and a training-free mechanism for answer reconciliation and selective answering in endoscopic VQA.

Figures

Figures reproduced from arXiv: 2607.17834 by Cheng Zhao, Guanghui Yue, Yuhao Liu.

Figure 1
Figure 1. Figure 1: Paired complex-atomic evaluation. A correct complex answer can still [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: ASR workflow for revision and selective answering. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Accuracy–consistency radar chart for six representative VLMs on [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Question-complexity analysis on EndoCA-Diagnostic. All-atomic [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Direct, ASR-Revise, and ASR-Selective complex-answer accuracy [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

30 extracted references · 10 linked inside Pith

  1. [1]

    VQA: Visual question answering,

    S. Antol, A. Agrawal, J. Lu, M. Mitchell, D. Batra, C. L. Zitnick, and D. Parikh, “VQA: Visual question answering,” inProceedings of the IEEE International Conference on Computer Vision, 2015, pp. 2425–2433

    2015

  2. [2]

    Medical visual question answering: A survey,

    Z. Lin, D. Zhang, Q. Tao, D. Shi, G. Haffari, Q. Wu, M. He, and Z. Ge, “Medical visual question answering: A survey,”Artificial Intelligence in Medicine, vol. 143, p. 102611, 2023

    2023

  3. [3]

    EndoBench: A comprehensive evaluation of multi-modal large language models for endoscopy analysis,

    R. Liuet al., “EndoBench: A comprehensive evaluation of multi-modal large language models for endoscopy analysis,” 2025, arXiv:2505.23601

    arXiv 2025

  4. [4]

    Kvasir-VQA: A text-image pair gi tract dataset,

    S. Gautam, A. Storås, C. Midoglu, S. A. Hicks, V . Thambawita, P. Halvorsen, and M. A. Riegler, “Kvasir-VQA: A text-image pair gi tract dataset,” inProceedings of the First International Workshop on Vision-Language Models for Biomedical Applications, 2024

    2024

  5. [5]

    Kvasir-VQA-x1: A multimodal dataset for medical reasoning and robust MedVQA in gastrointestinal endoscopy,

    S. Gautam, M. A. Riegler, and P. Halvorsen, “Kvasir-VQA-x1: A multimodal dataset for medical reasoning and robust MedVQA in gastrointestinal endoscopy,” 2025, arXiv:2506.09958

    Pith/arXiv arXiv 2025

  6. [6]

    A dataset of clinically generated visual questions and answers about radiology images,

    J. J. Lau, S. Gayen, A. Ben Abacha, and D. Demner-Fushman, “A dataset of clinically generated visual questions and answers about radiology images,”Scientific Data, vol. 5, p. 180251, 2018

    2018

  7. [7]

    SLAKE: A semantically-labeled knowledge-enhanced dataset for medical visual question answering,

    B. Liu, L.-M. Zhan, L. Xu, L. Ma, Y . Yang, and X.-M. Wu, “SLAKE: A semantically-labeled knowledge-enhanced dataset for medical visual question answering,” in2021 IEEE 18th International Symposium on Biomedical Imaging, 2021

    2021

  8. [8]

    PathVQA: 30000+ ques- tions for medical visual question answering,

    X. He, Y . Zhang, L. Mou, E. P. Xing, and P. Xie, “PathVQA: 30000+ ques- tions for medical visual question answering,” 2020, arXiv:2003.10286

    Pith/arXiv arXiv 2020

  9. [9]

    Gastric-X: A multimodal multi-phase benchmark dataset for advancing vision-language models in gastric cancer analysis,

    S. Luet al., “Gastric-X: A multimodal multi-phase benchmark dataset for advancing vision-language models in gastric cancer analysis,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2026

    2026

  10. [10]

    Multimodal AI for gastrointestinal diagnostics: Tackling VQA in MEDVQA-GI 2025,

    P. K. Gaihre, A. More, R. Kusumba, M. Bhalerao, S. Sinha, C. Bhatt, G. Thakur, D. K. Yadav, and A. K. Jaiswal, “Multimodal AI for gastrointestinal diagnostics: Tackling VQA in MEDVQA-GI 2025,” 2025, arXiv:2507.14544

    Pith/arXiv arXiv 2025

  11. [11]

    SQuINTing at VQA models: Introspecting VQA models with sub-questions,

    R. R. Selvaraju, P. Tendulkar, D. Parikh, E. Horvitz, M. T. Ribeiro, B. Nushi, and E. Kamar, “SQuINTing at VQA models: Introspecting VQA models with sub-questions,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 10 003–10 011

    2020

  12. [12]

    GQA: A new dataset for real- world visual reasoning and compositional question answering,

    D. A. Hudson and C. D. Manning, “GQA: A new dataset for real- world visual reasoning and compositional question answering,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019

    2019

  13. [13]

    FActScore: Fine-grained atomic eval- uation of factual precision in long form text generation,

    S. Min, K. Krishna, X. Lyu, M. Lewis, W.-t. Yih, P. W. Koh, M. Iyyer, L. Zettlemoyer, and H. Hajishirzi, “FActScore: Fine-grained atomic eval- uation of factual precision in long form text generation,” inProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023, pp. 12 076–12 100

    2023

  14. [14]

    FaithScore: Fine-grained evaluations of hallucinations in large vision-language models,

    L. Jing, R. Li, Y . Chen, and X. Du, “FaithScore: Fine-grained evaluations of hallucinations in large vision-language models,” inFindings of the Association for Computational Linguistics: EMNLP 2024, 2024, pp. 5042–5063

    2024

  15. [15]

    Evaluating object hallucination in large vision-language models,

    Y . Li, Y . Du, K. Zhou, J. Wang, X. Zhao, and J.-R. Wen, “Evaluating object hallucination in large vision-language models,” inProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023, pp. 292–305

    2023

  16. [16]

    SurgViVQA: Temporally grounded video question answering for surgical scene understanding,

    M. O. Drago, L. Carlini, P. C. Balyemez, D. Pierantozzi, C. Lena, C. Hassan, D. Stoyanov, E. De Momi, S. Bano, and M. I. Hoque, “SurgViVQA: Temporally grounded video question answering for surgical scene understanding,”International Journal of Computer Assisted Radiology and Surgery, 2026

    2026

  17. [17]

    SOrT- ing VQA Models: Contrastive gradient learning for improved consistency,

    S. Dharur, P. Tendulkar, D. Batra, D. Parikh, and R. R. Selvaraju, “SOrT- ing VQA Models: Contrastive gradient learning for improved consistency,” inProceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2021, pp. 3103–3111

    2021

  18. [18]

    Self-consistency improves chain of thought reasoning in language models,

    X. Wang, J. Wei, D. Schuurmans, Q. Le, E. Chi, S. Narang, A. Chowdhery, and D. Zhou, “Self-consistency improves chain of thought reasoning in language models,” inInternational Conference on Learning Representa- tions, 2023

    2023

  19. [19]

    SelfCheckGPT: Zero-resource black-box hallucination detection for generative large language models,

    P. Manakul, A. Liusie, and M. J. F. Gales, “SelfCheckGPT: Zero-resource black-box hallucination detection for generative large language models,” inProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023, pp. 9004–9017

    2023

  20. [20]

    EndoCogniAgent: Closed-loop agentic reasoning with self-consistency validation for endoscopic diagnosis,

    S. Tanget al., “EndoCogniAgent: Closed-loop agentic reasoning with self-consistency validation for endoscopic diagnosis,” 2025, arXiv:2508.07292

    Pith/arXiv arXiv 2025

  21. [21]

    Expanding performance boundaries of open-source multimodal models with model, data, and test-time scaling,

    Z. Chenet al., “Expanding performance boundaries of open-source multimodal models with model, data, and test-time scaling,” 2024, arXiv:2412.05271

    Pith/arXiv arXiv 2024

  22. [22]

    Qwen2.5-VL technical report,

    S. Baiet al., “Qwen2.5-VL technical report,” 2025, arXiv:2502.13923

    Pith/arXiv arXiv 2025

  23. [23]

    Qwen3-VL Technical Report,

    S. Baiet al., “Qwen3-VL Technical Report,” 2025, arXiv:2511.21631

    Pith/arXiv arXiv 2025

  24. [24]

    MedGemma technical report,

    A. Sellergrenet al., “MedGemma technical report,” 2025, arXiv:2507.05201

    Pith/arXiv arXiv 2025

  25. [25]

    Lingshu: A generalist foundation model for unified multimodal medical understanding and reasoning,

    LASA Team, W. Xuet al., “Lingshu: A generalist foundation model for unified multimodal medical understanding and reasoning,” 2025, arXiv:2506.07044

    Pith/arXiv arXiv 2025

  26. [26]

    LLaV A-Med: Training a large language-and-vision assistant for biomedicine in one day,

    C. Li, C. Wong, S. Zhang, N. Usuyama, H. Liu, J. Yang, T. Naumann, H. Poon, and J. Gao, “LLaV A-Med: Training a large language-and-vision assistant for biomedicine in one day,” inAdvances in Neural Information Processing Systems, vol. 36, 2023, pp. 28 541–28 564, datasets and Benchmarks Track

    2023

  27. [27]

    LoRA: Low-rank adaptation of large language models,

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-rank adaptation of large language models,” in International Conference on Learning Representations, 2022. [Online]. Available: https://arxiv.org/abs/2106.09685

    Pith/arXiv arXiv 2022

  28. [28]

    Introducing GPT-5.5,

    OpenAI, “Introducing GPT-5.5,” https://openai.com/index/ introducing-gpt-5-5/, Apr. 2026

    2026

  29. [29]

    Grok 4.20,

    xAI, “Grok 4.20,” xAI Docs. https://docs.x.ai/developers/models/grok-4. 20, 2026

    2026

  30. [30]

    Introducing Claude Opus 4.7,

    Anthropic, “Introducing Claude Opus 4.7,” Anthropic News. https://www. anthropic.com/news/claude-opus-4-7, Apr. 2026

    2026