Pith. sign in

REVIEW 3 major objections 6 minor 2 cited by

Cross-Modal Consistency in Multimodal Large Language Models

T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read GPT-4V is cross-modally inconsistent: for task instances carrying the same information, its output depends on whether the input is text or image.

desk verdict Useful formalization of cross-modal consistency with a robust but thinly evidenced GPT-4V finding; deserves peer review with revisions. read the letter →

arxiv 2411.09273 v1 pith:XCWBSSIA submitted 2024-11-14 cs.CL cs.AI

classification cs.CLcs.AI
keywords cross-modalconsistencymultimodallargelanguagemodelsGPT-4Vvision-languageevaluationinformation-preservingconversionVision-Depicting-Promptingparalleldataset
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 tries to establish that GPT-4V, a multimodal model that handles both vision and language, does not treat equivalent information equally across modalities. The authors define cross-modal consistency as the requirement that a model give the same answer to a task instance whether it is presented as text or as an image, under an information-preserving conversion. Across seven parallel vision-language datasets they constructed, GPT-4V's accuracy drops substantially when the same problem arrives as an image, in one case from 0.93 in text to 0.03 in image. The paper concludes that GPT-4V has a measurable language-side bias in its internal reasoning, and proposes a two-step prompting method, Vision-Depicting-Prompting, that routes image content through text and recovers much of the gap.

What carries the argument

The framework rests on the notion of an information-preserving converter $K^q_{a,b}$ between modalities, and on the consistency score $C_t$ that measures how often outputs agree across converted pairs. The authors build parallel datasets for seven tasks using OCR with human verification for image-to-text conversion and rendered screenshots for text-to-image conversion, selecting tasks where they argue the conversion preserves all task-relevant information. The Vision-Depicting-Prompting (VDP) method is the corrective mechanism: it first asks the model to transcribe the image into text, then answers with both the transcription and the original image, exploiting the model's stronger text-side reasoning.

What would settle it

Run the seven parallel tasks on a fresh GPT-4V session with the same converters and manually verified conversions; if image-mode accuracy on Table Understanding and MMLU matches text-mode accuracy, or the consistency score $C_t$ exceeds 0.95 across tasks, then the reported cross-modal inconsistency would fail to replicate and the claim would be withdrawn.

Watch

Extended reading notes

Core claim

The central discovery is that GPT-4V is cross-modally inconsistent as formalized by the paper's consistency score $C_t = \frac{1}{n}\sum_i c_i^M$, the fraction of paired instances where the model's answer is identical under text and image input. Table 1 reports large image-mode accuracy drops on all seven tasks, including Table Understanding (0.93 text vs 0.03 image) and MMLU (1.00 vs 0.74), while a one-step OCR ablation shows near-perfect extraction on LogicQA, MMLU, and Table Understanding. Because the information in the image is available to the model, the authors attribute the gap to the model's internal reasoning processes differing by modality rather than to failed perception. The Vision-Depicting-Prompting results support the same interpretation: having the model first describe the image in text raises image-mode accuracy by up to 57 percentage points on understanding tasks and substantially increases consistency.

Load-bearing premise

The measured gaps rest on the assumption that the converters—OCR with human verification and rendered screenshots—preserve all task-relevant information, including layout, for every one of the seven tasks; the authors concede in the Limitations section that lossless conversion is not generally available, so a conversion artifact could mimic a true inconsistency.

Editorial extensions

If this is right

  • Users of GPT-4V should expect markedly worse answers on image-only inputs than on equivalent text inputs, especially for table and reading-comprehension tasks.
  • Cross-modal consistency should be reported alongside per-modality accuracy, since a model can score well in each modality taken alone yet be unreliable across them.
  • For tasks where OCR is nearly perfect, the bottleneck is not visual information extraction but modality-specific reasoning, so improving visual reasoning is the targeted next step.
  • VDP-style prompting is a practical workaround: on Table Understanding it raises image accuracy from 0.03 to 0.93 and consistency from 0.10 to 0.90.

Reading between the lines

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

  • The same converter-based consistency test could be applied to other multimodal models to determine whether a language-side bias is specific to GPT-4V or a general property of vision-language alignment.
  • If the bias is architectural, then input modality becomes a reliability surface: the same prompt could yield different factual behavior depending on whether it is delivered as text or as a screenshot.
  • A natural extension is to use consistency scores as a training objective or as a diagnostic signal for modality-invariant representations.
  • The framework could be extended to audio or video modalities whenever an information-preserving converter exists, testing whether the language-dominant pattern generalizes.
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

3 major / 6 minor

Summary. The paper defines cross-modal consistency for multimodal LLMs as output invariance under information-preserving modality conversion (Eq. 1), constructs parallel text-image datasets for seven tasks, evaluates GPT-4V on 50 instances per task, and reports substantial accuracy drops when input is presented as an image rather than text (e.g., Table Understanding 0.93 to 0.03, MMLU 1.00 to 0.74). It also proposes a Vision-Depicting-Prompting (VDP) method that extracts a textual description of the image before answering, which partially recovers the image-mode deficit and improves consistency scores. The central conclusion is that GPT-4V exhibits language-dominant, modality-dependent internal reasoning, rather than a unified multimodal understanding.

Significance. If the result is sound, the paper makes a useful contribution by formalizing a cross-modal consistency metric and providing a concrete evaluation framework. The dataset release (promised but not yet available) and the VDP method are potentially valuable. The paper is honest about the difficulty of information-preserving conversion and explicitly acknowledges limitations. However, the strength of the central claim depends on two premises: that the converters preserve all task-relevant information (including layout), and that the 50-instance subsets are representative. Both premises are currently insufficiently supported, so the evidence for 'solely internal reasoning' remains incomplete.

major comments (3)
  1. [Section 5.4, Tables 2-3] The ablation study does not rule out perceptual failure as the source of the image deficit. For Table Understanding, an OCR accuracy of 1.00 only shows that the model can transcribe the numbers; the task also requires locating row/column positions in the rendered image, which the OCR check does not verify. The conditional consistency scores in Table 3 are consistency rates, not conditional accuracies, so they cannot isolate whether the gap stems from failed spatial perception or from modality-dependent reasoning. The claim that the gap 'can be attributed solely to the model's internal reasoning processes' (Section 5.4) is therefore not supported by the reported evidence.
  2. [Section 5.2, Table 1] The evaluation uses 50 randomly selected instances per task with no confidence intervals, significance tests, or released data. The MMLU text accuracy of 1.00 suggests the sampled split is unusually easy and may not be representative. With n=50, the 95% confidence interval for a proportion is roughly ±0.14, so the 'sizeable decrease' threshold of 10% is within the noise band for several tasks. The paper should report bootstrap or exact confidence intervals and make the sampled instances public to substantiate the cross-task generalization claim.
  3. [Sections 3 and 4.1, Limitations] The information-preservation assumption is load-bearing for the entire framework, yet the Limitations section acknowledges that lossless conversion is not generally available. For Math Equation Solving, the OCR accuracy is only 0.68-0.76 (Table 2), so the converter demonstrably does not preserve information for a substantial fraction of instances. The conditional consistency analysis in Table 3 does not report task accuracy on the subset with correct OCR, so it cannot separate conversion artifacts from genuine cross-modal inconsistency. The authors should report image accuracy and text accuracy on the same subset of instances where OCR is verified correct, to establish that the gap persists even when information is actually preserved.
minor comments (6)
  1. [Figure 1] The figure label 'T ext' should read 'Text'; the figure is also not referenced in the body text before it appears.
  2. [Section 3] The notation for the converter is inconsistent: K^q_{a,b} is used in the formal definition, but K_{a,b} appears in Section 4.1. Clarify whether the converter is query-dependent in all uses.
  3. [Section 5.4] The statement that 'OCR accuracy approaches nearly 100% for all instances of LogicQA, MMLU, and Table Understanding tasks' should specify that this refers to the 50-image subsets used in the experiments, not the full source datasets.
  4. [Appendix A and B] The text versions of math equations contain unrendered LaTeX (e.g., 'x ˆ 2', '$$2x + 4 = |-17 + 3|$$'), which makes the appendix difficult to read; the intended rendered formulas should be shown.
  5. [Section 5.2] The paper states that 'our results are manually collected for pairwise data instances,' but no inter-annotator agreement or verification protocol is described; at least a brief description would strengthen reproducibility.
  6. [Ethical Consideration] The ethical consideration paragraph is generic and does not discuss the specific ethical implications of reporting consistency failures in a commercial model or the potential for misuse of the VDP method; this is a presentation issue rather than a technical one.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: cross-modal consistency is measured directly, not derived from its definition.

full rationale

The paper's central quantity is Eq. (1), an invariance definition: a model is consistent if its outputs agree on information-preserving modality-converted instances. This definition imposes no constraint on GPT-4V's actual outputs, so the reported inconsistencies are empirical findings obtained by directly querying the model on parallel text/image instances. The result therefore carries independent content and is not forced by construction. The only self-citation (Zhang et al. 2023b) motivates the framework by analogy to cross-lingual consistency and is not load-bearing for the GPT-4V measurement; it is not used as a uniqueness theorem or to forbid alternatives. The information-preserving converter assumption is a validity condition for interpreting the measurements, not a step that makes the conclusion true by definition; the paper's own Limitations explicitly concede that lossless conversion is not generally available, which is an honest statement of scope rather than a circular move. VDP is a post-hoc prompting intervention designed from the observed image/text gap and then evaluated on the same tasks; this is a normal empirical loop, not a prediction that reduces to its own inputs. No fitted parameter is renamed as a prediction, and no ansatz is smuggled in via self-citation. Thus the derivation chain is self-contained and the measured cross-modal inconsistency stands as an independent empirical claim.

Assumptions & free parameters 1 free parameters · 3 assumptions · 2 invented entities

The framework introduces no fitted numeric parameters; the only hand-chosen numbers are the 50-instance subset sizes, the low decoding temperature, and the 10% threshold used to annotate 'sizeable' drops in Table 1. The load-bearing premises are the existence and losslessness of the modality converters, exact string equality as the operational definition of answer agreement, and representativeness of the small subsets. No new physical entities are postulated; the invented constructs are the consistency definition and the VDP prompting method.

free parameters (1)
  • Performance-gap threshold for 'sizeable decrease' = 10%
    Table 1 uses a hand-chosen greater-than-10% accuracy drop to flag a sizeable decrease between text and image modalities. The threshold is interpretive and does not affect the computed consistency scores, but it shapes the narrative of the results.
assumptions (3)
  • domain assumption Existence of an information-preserving converter K_q(a,b) for each task, and that OCR-plus-human-verification and screenshot converters instantiate it losslessly for the chosen tasks.
    Section 3 postulates existence of K_q for every q; Sections 4.1 and 5.1 assert that OCR with human verification and screenshots are lossless for the selected tasks. If rendering or OCR loses task-relevant layout (e.g., table cell positions), the measured gap conflates conversion artifacts with cross-modal inconsistency.
  • domain assumption Exact string equality of model outputs is a valid operationalization of answer agreement.
    Equations (1) and (2) define consistency as M(da,q) = M(db,q) string-wise; mathematically equivalent answers with different surface forms (e.g., '1/2' vs '0.5') would be scored inconsistent, potentially deflating consistency scores.
  • domain assumption The 50-instance random subsets are representative of each source dataset, and low decoding temperature makes pairwise comparisons stable.
    Section 5.2 samples 50 instances per dataset due to API limits; no confidence intervals or repeat runs are reported, yet all conclusions are drawn from these single 50-sample draws.
invented entities (2)
  • Cross-modal consistency (formal definition)
    purpose: Primary evaluation construct: invariance of model output under information-preserving modality conversion (Eq. 1-2).
    A definition, not an observable entity; its operationalization depends on the converter axiom above. It is measurable in principle but has no falsifiable handle outside the paper beyond the datasets, which are not yet released.
  • Vision-Depicting-Prompting (VDP)
    purpose: Two-step prompting method: extract a textual description of the image, then answer from image plus description, to exploit the model's text bias.
    VDP's effectiveness is measured on the same five datasets whose modality gap motivated the method (Section 6.2, Table 4); no held-out or external benchmark validation is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cross-Modal Consistency in Multimodal Large Language Models." pith.science (2026). https://pith.science/paper/XCWBSSIA

@misc{pith2026241109273,
  author       = {Pith},
  title        = {Pith review of: Cross-Modal Consistency in Multimodal Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XCWBSSIA}},
  note         = {Machine review of arXiv:2411.09273}
}
read the original abstract

Recent developments in multimodal methodologies have marked the beginning of an exciting era for models adept at processing diverse data types, encompassing text, audio, and visual content. Models like GPT-4V, which merge computer vision with advanced language processing, exhibit extraordinary proficiency in handling intricate tasks that require a simultaneous understanding of both textual and visual information. Prior research efforts have meticulously evaluated the efficacy of these Vision Large Language Models (VLLMs) in various domains, including object detection, image captioning, and other related fields. However, existing analyses have often suffered from limitations, primarily centering on the isolated evaluation of each modality's performance while neglecting to explore their intricate cross-modal interactions. Specifically, the question of whether these models achieve the same level of accuracy when confronted with identical task instances across different modalities remains unanswered. In this study, we take the initiative to delve into the interaction and comparison among these modalities of interest by introducing a novel concept termed cross-modal consistency. Furthermore, we propose a quantitative evaluation framework founded on this concept. Our experimental findings, drawn from a curated collection of parallel vision-language datasets developed by us, unveil a pronounced inconsistency between the vision and language modalities within GPT-4V, despite its portrayal as a unified multimodal model. Our research yields insights into the appropriate utilization of such models and hints at potential avenues for enhancing their design.

Figures

Figures reproduced from arXiv: 2411.09273 by the authors.

Figure 1
Figure 1. Visualization of the performance gap between [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the concept of cross-modal con [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. An Overview of the Components of Our Vision-Language Consistency Dataset. Data instances are [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (15 more)
Figure 4
Figure 4. Figure 4: Overview of the VDP Method: The left part illustrates the conventional approach to prompting vision [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Sample 1 of Math Equation Solving (Easy) Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Sample 2 of Math Equation Solving (Easy) Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Sample 1 of Math Equation Solving (Hard) Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 8
Figure 8. Figure 8: Sample 2 of Math Equation Solving (Hard) Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Sample 1 of LogicQA Dataset: Image. Text: Give me only a single choice, NO EXPLANATIONS AT ALL! Choose only one choice from below. Which of the followings, if true, can best support the above statement? Given that jupiter is a gas giant planet and the largest planet in…
Figure 10
Figure 10. Figure 10: Sample 2 of LogicQA Dataset: Image. Text: Give me only a single choice, NO EXPLANATIONS AT ALL! Choose only one choice from below. Which of the followings can be infered Given that all Anxi people are vegetarians, while all Zhenyuan people are ascetics. Ascetics and v…
Figure 11
Figure 11. Figure 11: Sample 1 of MMLU Dataset: Image. Text: Give me only a single letter, NO EXPLANATIONS AT ALL! Choose one from below. Tom had to fix some things around the house. He had to fix the door. He had to fix the window. But before he did anything he had to fix the toilet. Tom …
Figure 12
Figure 12. Figure 12: Sample 2 of MMLU Dataset: Image. Text: Give me only a single letter, NO EXPLANATIONS AT ALL! Choose one from below. Lisa has a pet cat named Whiskers. Whiskers is black with a white spot on her chest. Whiskers also has white paws that look like little white mittens. W…
Figure 13
Figure 13. Figure 13: Sample 1 of Table Understanding Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: Sample 2 of Table Understanding Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p020_14.png]
Figure 15
Figure 15. Figure 15: Sample 1 of Math Reasoning Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p021_15.png]
Figure 16
Figure 16. Figure 16: Sample 2 of Math Reasoning Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p022_16.png]
Figure 17
Figure 17. Figure 17: Sample 1 of State Machine Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p023_17.png]
Figure 18
Figure 18. Figure 18: Sample 2 of State Machine Dataset: Image. [PITH_FULL_IMAGE:figures/full_fig_p024_18.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. TokenSwap: Benchmarking and Reducing the Modality Gap in Multimodal LLMs

    cs.CL 2026-05 conditional novelty 7.0 of 10

    TokenSwap measures and mitigates the MLLM modality gap: swapping textual concepts for matched images lowers accuracy by 4-47% across 42 models, and training with such swaps reduces the gap.

  2. SEAM: Semantically Equivalent Across Modalities Benchmark for Vision-Language Models

    cs.AI 2025-08 unverdicted novelty 6.0 of 10

    SEAM measures VLM reasoning consistency across modalities using paired semantically equivalent textual and visual notations, and finds systematic vision-language imbalance.

Reference graph

Works this paper leans on

24 extracted references · 4 canonical work pages · cited by 2 Pith papers

  1. [1]

    Liang Chen, Yichi Zhang, Shuhuai Ren, Haozhe Zhao, Zefan Cai, Yuchi Wang, Peiyi Wang, Tianyu Liu, and Baobao Chang. 2023. http://arxiv.org/abs/2310.02071 Towards end-to-end embodied decision making via multi-modal large language model: Explorations with gpt4-vision and beyond

  2. [2]

    Shih-Han Chou, Shivam Chandhok, James J Little, and Leonid Sigal. 2024. Mm-r 3: On (in-) consistency of multi-modal large language models (mllms). arXiv preprint arXiv:2410.04778

  3. [3]

    Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. https://arxiv.org/abs/2110.14168 Training verifiers to solve math word problems . CoRR

  4. [4]

    Danny Driess, Fei Xia, Mehdi S. M. Sajjadi, Corey Lynch, Aakanksha Chowdhery, Brian Ichter, Ayzaan Wahid, Jonathan Tompson, Quan Vuong, Tianhe Yu, Wenlong Huang, Yevgen Chebotar, Pierre Sermanet, Daniel Duckworth, Sergey Levine, Vincent Vanhoucke, Karol Hausman, Marc Toussaint, Klaus Greff, Andy Zeng, Igor Mordatch, and Pete Florence. 2023. http://arxiv.o...

  5. [5]

    Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021 a . http://arxiv.org/abs/2009.03300 Measuring massive multitask language understanding

  6. [6]

    Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021 b . http://arxiv.org/abs/2103.03874 Measuring mathematical problem solving with the math dataset

  7. [7]

    Yushi Hu, Hang Hua, Zhengyuan Yang, Weijia Shi, Noah A Smith, and Jiebo Luo. 2023. http://arxiv.org/abs/2211.09699 Promptcap: Prompt-guided task-aware image captioning

  8. [8]

    Hanyao Huang, Ou Zheng, Dongdong Wang, Jiayi Yin, Zijin Wang, Shengxuan Ding, Heng Yin, Chuan Xu, Renjie Yang, Qian Zheng, et al. 2023. Chatgpt for shaping the future of dentistry: the potential of multi-modal large language model. International Journal of Oral Science, 15(1):29

Show all 24 references
  1. [9]

    Zhi Jin, Sheng Xu, Xiang Zhang, Tianze Ling, Nanqing Dong, Wanli Ouyang, Zhiqiang Gao, Cheng Chang, and Siqi Sun. 2024. Contranovo: A contrastive learning approach to enhance de novo peptide sequencing. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 3...

  2. [10]

    Yuanze Lin, Yujia Xie, Dongdong Chen, Yichong Xu, Chenguang Zhu, and Lu Yuan. 2022. http://arxiv.org/abs/2206.01201 Revive: Regional visual representation matters in knowledge-based visual question answering

  3. [11]

    Fuxiao Liu, Xiaoyang Wang, Wenlin Yao, Jianshu Chen, Kaiqiang Song, Sangwoo Cho, Yaser Yacoob, and Dong Yu. 2023 a . http://arxiv.org/abs/2311.10774 Mmc: Advancing multimodal chart understanding with large-scale instruction tuning

  4. [12]

    Jian Liu, Leyang Cui, Hanmeng Liu, Dandan Huang, Yile Wang, and Yue Zhang. 2020. http://arxiv.org/abs/2007.08124 Logiqa: A challenge dataset for machine reading comprehension with logical reasoning

  5. [13]

    Zhengliang Liu, Hanqi Jiang, Tianyang Zhong, Zihao Wu, Chong Ma, Yiwei Li, Xiaowei Yu, Yutong Zhang, Yi Pan, Peng Shu, Yanjun Lyu, Lu Zhang, Junjie Yao, Peixin Dong, Chao Cao, Zhenxiang Xiao, Jiaqi Wang, Huan Zhao, Shaochen Xu, Yaonai Wei, Jingyuan Chen, Haixing Dai, Peilong W...

  6. [14]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. http://arxiv.org/abs/2103.00020 Learning transferable visual models from natural lang...

  7. [15]

    Yongxin Shi, Dezhi Peng, Wenhui Liao, Zening Lin, Xinhong Chen, Chongyu Liu, Yuyi Zhang, and Lianwen Jin. 2023. http://arxiv.org/abs/2310.16809 Exploring ocr capabilities of gpt-4v(ision) : A quantitative and in-depth evaluation

  8. [16]

    Licheng Wen, Xuemeng Yang, Daocheng Fu, Xiaofeng Wang, Pinlong Cai, Xin Li, Tao Ma, Yingxuan Li, Linran Xu, Dengke Shang, Zheng Zhu, Shaoyan Sun, Yeqi Bai, Xinyu Cai, Min Dou, Shuanglu Hu, Botian Shi, and Yu Qiao. 2023. http://arxiv.org/abs/2311.05332 On the road with gpt-4v(i...

  9. [17]

    Chaoyi Wu, Jiayu Lei, Qiaoyu Zheng, Weike Zhao, Weixiong Lin, Xiaoman Zhang, Xiao Zhou, Ziheng Zhao, Ya Zhang, Yanfeng Wang, and Weidi Xie. 2023. http://arxiv.org/abs/2310.09909 Can gpt-4v(ision) serve medical applications? case studies on gpt-4v for multimodal medical diagnosis

  10. [18]

    Zhengyuan Yang, Linjie Li, Kevin Lin, Jianfeng Wang, Chung-Ching Lin, Zicheng Liu, and Lijuan Wang. 2023. The dawn of lmms: Preliminary explorations with gpt-4v (ision). arXiv preprint arXiv:2309.17421

  11. [19]

    Yuwei Yin, Jean Kaddour, Xiang Zhang, Yixin Nie, Zhenguang Liu, Lingpeng Kong, and Qi Liu. 2023. http://arxiv.org/abs/2304.08821 Ttida: Controllable generative data augmentation via text-to-text and text-to-image models

  12. [20]

    Jingyi Zhang, Jiaxing Huang, Sheng Jin, and Shijian Lu. 2023 a . Vision-language models for vision tasks: A survey. arXiv preprint arXiv:2304.00685

  13. [21]

    Xiang Zhang, Senyu Li, Bradley Hauer, Ning Shi, and Grzegorz Kondrak. 2023 b . Don't trust gpt when your question is not in english. arXiv preprint arXiv:2305.16339

  14. [22]

    Xiang Zhang, Ning Shi, Bradley Hauer, and Grzegorz Kondrak. 2023 c . https://doi.org/10.18653/v1/2023.eacl-main.205 Bridging the gap between B abel N et and H ow N et: Unsupervised sense alignment and sememe prediction . In Proceedings of the 17th Conference of the European Ch...

  15. [23]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before...

  16. [24]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

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