---
title: "Automatic and Reliable Evaluation for Academic"
description: "The paper also introduces two models: HE4AFG-E is an evaluator built on a scientifically adapted CLIP (SCLIP). 53 references."
authors: ["Guanghui Ye", "Huan Zhao", "Qin Zhu", "Fengnan Li", "Jiaqi Li", "Yixian Shen", "Zhonghao Ren", "Zhihua Jiang"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)"
source: "https://aclanthology.org/2026.acl-long.2055/"
doi: "10.18653/v1/2026.acl-long.2055"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/ye-2026-automatic-reliable-evaluation-academic-caption/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/ye-2026-automatic-reliable-evaluation-academic-caption/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/23/) / Automatic and Reliable Evaluation for Academic Caption-to-Figure Generation with LMMs

# Automatic and Reliable Evaluation for Academic Caption-to-Figure Generation with LMMs

**Authors:** Guanghui Ye, [Huan Zhao](https://md.research.elamur.ai/free/en/authors/huan-zhao/), [Qin Zhu](https://md.research.elamur.ai/free/en/authors/qin-zhu/), Fengnan Li, [Jiaqi Li](https://md.research.elamur.ai/free/en/authors/jiaqi-li/), Yixian Shen, Zhonghao Ren and [Zhihua Jiang](https://md.research.elamur.ai/free/en/authors/zhihua-jiang/) · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-long.2055/) · [PDF](https://aclanthology.org/2026.acl-long.2055.pdf) · [DOI 10.18653/v1/2026.acl-long.2055](https://doi.org/10.18653/v1/2026.acl-long.2055) · [OpenAlex](https://openalex.org/W7166827875)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Benchmarks for text-to-image generation mostly target everyday photographs, so they are poorly suited to judging whether a model can turn a real scientific caption into a correct academic figure. The authors address this with HE4AFG, a dataset for Academic caption-to-Figure Generation (AFG). They gathered 650 caption-figure pairs from arXiv papers and SciCap across 8 domains, about 10% of them multi-panel figures. Five generative LMMs then produced additional figures from these captions, giving 3,900 samples in total. Experts rated every sample along three dimensions: how well it matches the topic (TR), whether its attributes are correct (AC), and how scientifically aesthetic it looks (SA).

The paper also introduces two models:
- **HE4AFG-E** is an evaluator built on a scientifically adapted CLIP (SCLIP). It is trained with intra- and cross-modal contrastive learning on aspect-specific hard negatives derived from ArXivCap.
- **HE4AFG-R** fine-tunes a lightweight LMM on quality feedback and uses it to refine figures iteratively.

HE4AFG-E agrees with human ratings more closely than metric baselines and frontier LMMs, including GPT-4o, while being far cheaper to run. This makes automatic, fine-grained evaluation of scientific figure generation practical.

## Key points
- HE4AFG contains 3,900 samples built from 650 real captions across 8 arXiv domains: cs, econ, q-fin, q-bio, eess, stat, physics and math. The figures come from GPT-5, Qwen3, DALL-E, Llama and Stable Diffusion 3.5, and each sample has human SA/TR/AC scores.
- HE4AFG-E trains three aspect-specific sub-evaluators on CLIP encoders, using positive and negative pairs constructed from ArXivCap. Its training data does not overlap with the evaluation data.
- HE4AFG-E's correlation with human judgments is about 3 points higher than GPT-4o's on average. On the 650 original pairs it reaches 65.9/63.0/68.3 (TR/AC/SA), against 62.5/60.7/65.6 for GPT-4o. TIFA, CLIPScore and InstructBLIP-13b score far lower.
- Among the generators, human raters scored GPT-5 highest (TR 3.42, AC 3.10) and SD 3.5-Medium lowest (2.83, 2.43). The evaluator also generalizes to Q-Eval-100K and ScImage.
- The authors list two limitations:
  - The negatives are synthetic and fairly "idealized" compared with real model errors.
  - Multi-panel evaluation depends on how accurately the figures are segmented into sub-figures.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [University of Amsterdam](https://md.research.elamur.ai/free/en/universities/university-of-amsterdam/) — University, Amsterdam, Netherlands · 151 works · 174,506 citations
- [Duke University](https://md.research.elamur.ai/free/en/universities/duke-university/) — University, Durham, United States · 37 works · 15,861 citations
- [University of Warwick](https://md.research.elamur.ai/free/en/universities/university-of-warwick/) — University, Coventry, United Kingdom · 8 works · 1,291 citations
- [Jinan University](https://md.research.elamur.ai/free/en/universities/jinan-university/) — University, Guangzhou, China · 7 works · 91 citations
- [Hunan University](https://md.research.elamur.ai/free/en/universities/hunan-university/) — University, Changsha, China · 4 works · 304 citations

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 3,486 works

🧭 Topics (1): [📏 Evaluation & benchmarks](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/) · 1,341 works

## Cite

Cite as: Guanghui Ye, Huan Zhao, Qin Zhu, Fengnan Li, Jiaqi Li, Yixian Shen, Zhonghao Ren and Zhihua Jiang. 2026. Automatic and Reliable Evaluation for Academic Caption-to-Figure Generation with LMMs. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{ye2026automatic,
  title = {{Automatic and Reliable Evaluation for Academic Caption-to-Figure Generation with LMMs}},
  author = {Guanghui Ye, Huan Zhao, Qin Zhu, Fengnan Li, Jiaqi Li, Yixian Shen, Zhonghao Ren and Zhihua Jiang},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.2055},
  url = {https://doi.org/10.18653/v1/2026.acl-long.2055}
}
```

- [← previous Automatic Analysis of Collaboration through Human Conversational Data Resources: A Review](https://md.research.elamur.ai/free/en/papers/yu-2026-automatic-analysis-collaboration-through-human/)
- [next → Automatic Annotation of Mental Health Recovery Narratives: A Benchmark Study](https://md.research.elamur.ai/free/en/papers/pandey-2026-automatic-annotation-mental-health-recovery/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/23/)

## References

53 references, 22 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Rohan Anil, Sebastian Borgeaud, Yonghui Wu, Jean-Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M. Dai, Anja Hauth, Katie Millican, David Silver, Slav Petrov, Melvin Johnson, Ioannis Antonoglou, Julian Schrittwieser, Amelia Glaese, Jilin Chen, Emily Pitler, Timothy P. Lillicrap,… — [Gemini: A Family of Highly Capable Multimodal Models](https://md.research.elamur.ai/free/en/papers/gemini-team-2023-gemini-family-highly-capable-multimodal/)
2. Anthropic. 2024. Introducing the next generation of claude. [find](https://scholar.google.com/scholar?q=Anthropic%2E+2024%2E+Introducing+the+next+generation+of+claude%2E)
3. Jinze Bai, Shuai Bai, Shusheng Yang, Shijie Wang, Sinan Tan, Peng Wang, Junyang Lin, Chang Zhou, and Jingren Zhou. 2023. [Qwen-vl: A frontier large vision-language model with versatile abilities](https://md.research.elamur.ai/free/en/papers/bai-2023-qwen-vl-versatile-vision-language/). CoRR, abs/2308.12966.
4. Jonas Belouadi, Anne Lauscher, and Steffen Eger. 2024. [Automatikz: Text-guided synthesis of scientific vector graphics with tikz](https://md.research.elamur.ai/free/en/papers/belouadi-2023-automatikz-text-guided-synthesis-scientific/). In Proceedings of the International Conference on Learning Representations.
5. Iz Beltagy, Kyle Lo, and Arman Cohan. 2019. [Scibert: A pretrained language model for scientific text](https://md.research.elamur.ai/free/en/papers/beltagy-2019-scibert-pretrained-language-model-scientific/). In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, EMNLP-IJCNLP 2019, Hong Kong, Chi…
6. Jaemin Cho, Abhay Zala, and Mohit Bansal. 2023. DALL-EVAL: probing the reasoning skills and social biases of text-to-image generation models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3020–3031. [find](https://scholar.google.com/scholar?q=Jaemin+Cho%2C+Abhay+Zala%2C+and+Mohit+Bansal%2E+2023%2E+DALL%2DEVAL%3A+probing+the+reasoning+skills+and+social+biases+of+text%2Dto%2Dimage+generation+models%2E+In+Proceedings+of+the+IEEE%2FCVF+Conference+on+Computer+Visi)
7. Wenliang Dai, Junnan Li, Dongxu Li, Anthony Meng Huat Tiong, Junqi Zhao, Weisheng Wang, Boyang Li, Pascale Fung, and Steven C. H. Hoi. 2023. [Instructblip: Towards general-purpose vision-language models with instruction tuning](https://md.research.elamur.ai/free/en/papers/dai-2023-instructblip-towards-general-purpose-vision/). In Advances in Neural Information Processing Systems 36: Annual Conferenc…
8. Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jégou. 2024. [The faiss library](https://md.research.elamur.ai/free/en/papers/douze-2024-faiss-library/). CoRR, abs/2401.08281.
9. Jennifer D’Souza, Hamed Babaei Giglou, and Quentin Münch. 2025. Yescieval: Robust llm-as-a-judge for scientific question answering. In Proceedings of the Annual Meeting of the Association for Computational Linguistics, pages 13749–13783. [find](https://scholar.google.com/scholar?q=Jennifer+D%E2%80%99Souza%2C+Hamed+Babaei+Giglou%2C+and+Quentin+M%C3%BCnch%2E+2025%2E+Yescieval%3A+Robust+llm%2Das%2Da%2Djudge+for+scientific+question+answering%2E+In+Proceedings+of+the+Annual+Meeting+of+the+Association+for+Computat)
10. Yuying Ge, Sijie Zhao, Jinguo Zhu, Yixiao Ge, Kun Yi, Lin Song, Chen Li, Xiaohan Ding, and Ying Shan. 2024. SEED-X: multimodal models with unified multi-granularity comprehension and generation. CoRR, abs/2404.14396. — [arXiv:2404.14396](https://arxiv.org/abs/2404.14396) [find](https://scholar.google.com/scholar?q=Yuying+Ge%2C+Sijie+Zhao%2C+Jinguo+Zhu%2C+Yixiao+Ge%2C+Kun+Yi%2C+Lin+Song%2C+Chen+Li%2C+Xiaohan+Ding%2C+and+Ying+Shan%2E+2024%2E+SEED%2DX%3A+multimodal+models+with+unified+multi%2Dgranularity+comprehension+and+generation%2E+CoRR%2C)
11. Jack Hessel, Ari Holtzman, Maxwell Forbes, Ronan Le Bras, and Yejin Choi. 2021. [Clipscore: A reference-free evaluation metric for image captioning](https://md.research.elamur.ai/free/en/papers/hessel-2021-clipscore-reference-free-evaluation-metric/). In Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 7514–7528.
12. Kyota Higa, Masahiro Yamaguchi, and Toshinori Hosoi. 2023. ICCL: self-supervised intra- and cross-modal contrastive learning with 2d-3d pairs for 3d scene understanding. In Proceedings of International Conference on Image Processing, pages 1085–1089. [find](https://scholar.google.com/scholar?q=Kyota+Higa%2C+Masahiro+Yamaguchi%2C+and+Toshinori+Hosoi%2E+2023%2E+ICCL%3A+self%2Dsupervised+intra%2D+and+cross%2Dmodal+contrastive+learning+with+2d%2D3d+pairs+for+3d+scene+understanding%2E+In+Proceedings+of+Internationa)
13. Ting-Yao Hsu, C. Lee Giles, and Ting-Hao Kenneth Huang. 2021. Scicap: Generating captions for scientific figures. In Proceedings of Findings of the Conference on Empirical Methods in Natural Language Processing, pages 3258–3264. [find](https://scholar.google.com/scholar?q=Ting%2DYao+Hsu%2C+C%2E+Lee+Giles%2C+and+Ting%2DHao+Kenneth+Huang%2E+2021%2E+Scicap%3A+Generating+captions+for+scientific+figures%2E+In+Proceedings+of+Findings+of+the+Conference+on+Empirical+Methods+in+Natural+Language+)
14. Yushi Hu, Benlin Liu, Jungo Kasai, Yizhong Wang, Mari Ostendorf, Ranjay Krishna, and Noah A. Smith. 2023. [TIFA: accurate and interpretable text-to-image faithfulness evaluation with question answering](https://md.research.elamur.ai/free/en/papers/hu-2023-tifa-accurate-interpretable-text-image/). In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20349–…
15. Anil K. Jain, Karthik Nandakumar, and Arun Ross. 2005. Score normalization in multimodal biometric systems. Pattern Recognition, 38(12):2270–2285. [find](https://scholar.google.com/scholar?q=Anil+K%2E+Jain%2C+Karthik+Nandakumar%2C+and+Arun+Ross%2E+2005%2E+Score+normalization+in+multimodal+biometric+systems%2E+Pattern+Recognition%2C+38%2812%29%3A2270%E2%80%932285%2E)
16. Paul Kassianik, Baturay Saglam, Alexander Chen, Blaine Nelson, Anu Vellore, Massimo Aufiero, Fraser Burch, Dhruv Kedia, Avi Zohary, Sajana Weerawardhena, Aman Priyanshu, Adam Swanda, Amy Chang, Hyrum S. Anderson, Kojin Oshiba, Omar Santos, Yaron Singer, and Amin Karbasi. 2025. Llama-3.1-foundationai… — [arXiv:2504.21039](https://arxiv.org/abs/2504.21039) [find](https://scholar.google.com/scholar?q=Paul+Kassianik%2C+Baturay+Saglam%2C+Alexander+Chen%2C+Blaine+Nelson%2C+Anu+Vellore%2C+Massimo+Aufiero%2C+Fraser+Burch%2C+Dhruv+Kedia%2C+Avi+Zohary%2C+Sajana+Weerawardhena%2C+Aman+Priyanshu%2C+Adam+Swanda%2C+Amy+Chang%2C+Hyrum+)
17. Yuval Kirstain, Adam Polyak, Uriel Singer, Shahbuland Matiana, Joe Penna, and Omer Levy. 2023. Pick-a-pic: An open dataset of user preferences for text-to-image generation. In Proceedings of Advances in Neural Information Processing Systems. [find](https://scholar.google.com/scholar?q=Yuval+Kirstain%2C+Adam+Polyak%2C+Uriel+Singer%2C+Shahbuland+Matiana%2C+Joe+Penna%2C+and+Omer+Levy%2E+2023%2E+Pick%2Da%2Dpic%3A+An+open+dataset+of+user+preferences+for+text%2Dto%2Dimage+generation%2E+In+Proceedings+of+Advances+)
18. Hugo Laurençon, Lucile Saulnier, Léo Tronchon, Stas Bekman, Amanpreet Singh, Anton Lozhkov, Thomas Wang, Siddharth Karamcheti, Alexander M. Rush, Douwe Kiela, Matthieu Cord, and Victor Sanh. 2023. [OBELICS: an open web-scale filtered dataset of interleaved image-text documents](https://md.research.elamur.ai/free/en/papers/lauren-ccon-2023-obelics-open-web-scale-filtered/). In Advances in Neural…
19. Jialuo Li, Wenhao Chai, Xingyu Fu, Haiyang Xu, and Saining Xie. 2025. Science-t2i: Addressing scientific illusions in image synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2734–2744. [find](https://scholar.google.com/scholar?q=Jialuo+Li%2C+Wenhao+Chai%2C+Xingyu+Fu%2C+Haiyang+Xu%2C+and+Saining+Xie%2E+2025%2E+Science%2Dt2i%3A+Addressing+scientific+illusions+in+image+synthesis%2E+In+Proceedings+of+the+IEEE%2FCVF+Conference+on+Computer+Vision+and+)
20. Junnan Li, Dongxu Li, Silvio Savarese, and Steven C. H. Hoi. 2023. [BLIP-2: bootstrapping language-image pre-training with frozen image encoders and large language models](https://md.research.elamur.ai/free/en/papers/li-2023-blip-2-bootstrapping-language-image/). In International Conference on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA, volume 202 of Proceedings of…
21. Lei Li, Yuqi Wang, Runxin Xu, Peiyi Wang, Xiachong Feng, Lingpeng Kong, and Qi Liu. 2024. [Multimodal arxiv: A dataset for improving scientific comprehension of large vision-language models](https://md.research.elamur.ai/free/en/papers/li-2024-multimodal-arxiv-dataset-improving-scientific/). In Proceedings of the Annual Meeting of the Association for Computational Linguistics, pages 14369–14387.
22. Zhiqiu Lin, Deepak Pathak, Baiqi Li, Jiayao Li, Xide Xia, Graham Neubig, Pengchuan Zhang, and Deva Ramanan. 2024. Evaluating text-to-visual generation with image-to-text generation. In Computer Vision - ECCV 2024 - 18th European Conference, Milan, Italy, September 29-October 4, 2024, Proceedings, Pa… [find](https://scholar.google.com/scholar?q=Zhiqiu+Lin%2C+Deepak+Pathak%2C+Baiqi+Li%2C+Jiayao+Li%2C+Xide+Xia%2C+Graham+Neubig%2C+Pengchuan+Zhang%2C+and+Deva+Ramanan%2E+2024%2E+Evaluating+text%2Dto%2Dvisual+generation+with+image%2Dto%2Dtext+generation%2E+In+Computer+Vision)
23. Haotian Liu, Chunyuan Li, Yuheng Li, and Yong Jae Lee. 2024. [Improved baselines with visual instruction tuning](https://md.research.elamur.ai/free/en/papers/liu-2023-improved-baselines-visual-instruction-tuning/). In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22, 2024, pages 26286–26296. IEEE.
24. OpenAI. 2023a. Gpt-4v(ision) system card. [find](https://scholar.google.com/scholar?q=OpenAI%2E+2023a%2E+Gpt%2D4v%28ision%29+system+card%2E)
25. OpenAI. 2023b. New models and developer products announced at devday. [find](https://scholar.google.com/scholar?q=OpenAI%2E+2023b%2E+New+models+and+developer+products+announced+at+devday%2E)
26. OpenAI. 2024. Hello gpt-4o. [find](https://scholar.google.com/scholar?q=OpenAI%2E+2024%2E+Hello+gpt%2D4o%2E)
27. OpenAI. 2025. Gpt-5 system card. [find](https://scholar.google.com/scholar?q=OpenAI%2E+2025%2E+Gpt%2D5+system+card%2E)
28. Haoyi Qiu, Wenbo Hu, Zi-Yi Dou, and Nanyun Peng. 2024. VALOR-EVAL: holistic coverage and faithfulness evaluation of large vision-language models. In Proceedings of Findings of the Association for Computational Linguistics, pages 1783–1805. [find](https://scholar.google.com/scholar?q=Haoyi+Qiu%2C+Wenbo+Hu%2C+Zi%2DYi+Dou%2C+and+Nanyun+Peng%2E+2024%2E+VALOR%2DEVAL%3A+holistic+coverage+and+faithfulness+evaluation+of+large+vision%2Dlanguage+models%2E+In+Proceedings+of+Findings+of+the+Association+for+Comp)
29. QwenTeam. 2025. [Qwen3 technical report](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/). CoRR, abs/2505.09388.
30. 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. [Learning transferable visual models from natural language supervision](https://md.research.elamur.ai/free/en/papers/radford-2021-learning-transferable-visual-models-natural/). In Proceedings of the International C…
31. Machel Reid, Nikolay Savinov, Denis Teplyashin, Dmitry Lepikhin, Timothy P. Lillicrap, Jean-Baptiste Alayrac, Radu Soricut, Angeliki Lazaridou, Orhan Firat, Julian Schrittwieser, Ioannis Antonoglou, Rohan Anil, Sebastian Borgeaud, Andrew M. Dai, Katie Millican, Ethan Dyer, Mia Glaese, Thibault Sotti… — [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://md.research.elamur.ai/free/en/papers/gemini-team-2024-gemini-1-5-unlocking-multimodal/)
32. Jonathan Roberts, Kai Han, Neil Houlsby, and Samuel Albanie. 2024. Scifibench: Benchmarking large multimodal models for scientific figure interpretation. [find](https://scholar.google.com/scholar?q=Jonathan+Roberts%2C+Kai+Han%2C+Neil+Houlsby%2C+and+Samuel+Albanie%2E+2024%2E+Scifibench%3A+Benchmarking+large+multimodal+models+for+scientific+figure+interpretation%2E)
33. In Proceedings of Advances in Neural Information Processing Systems. [find](https://scholar.google.com/scholar?q=In+Proceedings+of+Advances+in+Neural+Information+Processing+Systems%2E)
34. Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. [High-resolution image synthesis with latent diffusion models](https://md.research.elamur.ai/free/en/papers/rombach-2021-high-resolution-image-synthesis-latent/). In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022, pages 10674–10685. IEEE.
35. Michael Saxon, Fatima Jahara, Mahsa Khoshnoodi, Yujie Lu, Aditya Sharma, and William Yang Wang. 2024. Who evaluates the evaluations? objectively scoring text-to-image prompt coherence metrics with t2iscorescore (TS2). In Proceedings of Advances in Neural Information Processing Systems. [find](https://scholar.google.com/scholar?q=Michael+Saxon%2C+Fatima+Jahara%2C+Mahsa+Khoshnoodi%2C+Yujie+Lu%2C+Aditya+Sharma%2C+and+William+Yang+Wang%2E+2024%2E+Who+evaluates+the+evaluations%3F+objectively+scoring+text%2Dto%2Dimage+prompt+coherence+metrics+with+t2i)
36. Haoyang Su, Renqi Chen, Shixiang Tang, Zhenfei Yin, Xinzhe Zheng, Jinzhe Li, Biqing Qi, Qi Wu, Hui Li, Wanli Ouyang, Philip Torr, Bowen Zhou, and Nanqing Dong. 2025. [Many heads are better than one: Improved scientific idea generation by A llm-based multi-agent system](https://md.research.elamur.ai/free/en/papers/su-2025-many-heads-are-better-than/). In Proceedings of the Annual Me…
37. Quan Sun, Yufeng Cui, Xiaosong Zhang, Fan Zhang, Qiying Yu, Yueze Wang, Yongming Rao, Jingjing Liu, Tiejun Huang, and Xinlong Wang. 2024. Generative multimodal models are in-context learners. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, Seattle, WA, USA, June 16-22,… [find](https://scholar.google.com/scholar?q=Quan+Sun%2C+Yufeng+Cui%2C+Xiaosong+Zhang%2C+Fan+Zhang%2C+Qiying+Yu%2C+Yueze+Wang%2C+Yongming+Rao%2C+Jingjing+Liu%2C+Tiejun+Huang%2C+and+Xinlong+Wang%2E+2024%2E+Generative+multimodal+models+are+in%2Dcontext+learners%2E+In+IEEE%2F)
38. InternLM Team. 2023. Internlm: A multilingual language model with progressively enhanced capabilities. [find](https://scholar.google.com/scholar?q=InternLM+Team%2E+2023%2E+Internlm%3A+A+multilingual+language+model+with+progressively+enhanced+capabilities%2E)
39. Ultralytics. 2023. Yolov8. [https://github.com/](https://github.com/) ultralytics/ultralytics. Accessed: 2026-04-15. [↗](https://github.com/)
40. Aäron van den Oord, Yazhe Li, and Oriol Vinyals. 2018. [Representation learning with contrastive predictive coding](https://md.research.elamur.ai/free/en/papers/oord-2018-representation-learning-contrastive-predictive-coding/). CoRR, abs/1807.03748.
41. Zirui Wang, Mengzhou Xia, Luxi He, Howard Chen, Yitao Liu, Richard Zhu, Kaiqu Liang, Xindi Wu, Haotian Liu, Sadhika Malladi, Alexis Chevalier, Sanjeev Arora, and Danqi Chen. 2024. [Charxiv: Charting gaps in realistic chart understanding in multimodal llms](https://md.research.elamur.ai/free/en/papers/wang-2024-charxiv-charting-gaps-realistic-chart/). In Proceedings of Advances in Neural Informa…
42. Yue Wu, Jiaming Liu, Maoguo Gong, Peiran Gong, Xiaolong Fan, A. Kai Qin, Qiguang Miao, and Wenping Ma. 2024. Self-supervised intra-modal and cross-modal contrastive learning for point cloud understanding. IEEE Trans. Multim., 26:1626–1638. [find](https://scholar.google.com/scholar?q=Yue+Wu%2C+Jiaming+Liu%2C+Maoguo+Gong%2C+Peiran+Gong%2C+Xiaolong+Fan%2C+A%2E+Kai+Qin%2C+Qiguang+Miao%2C+and+Wenping+Ma%2E+2024%2E+Self%2Dsupervised+intra%2Dmodal+and+cross%2Dmodal+contrastive+learning+for+point+cloud+understand)
43. Tianyi Xiong, Xiyao Wang, Dong Guo, Qinghao Ye, Haoqi Fan, Quanquan Gu, Heng Huang, and Chunyuan Li. 2025. Llava-critic: Learning to evaluate multimodal models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13618–13628. [find](https://scholar.google.com/scholar?q=Tianyi+Xiong%2C+Xiyao+Wang%2C+Dong+Guo%2C+Qinghao+Ye%2C+Haoqi+Fan%2C+Quanquan+Gu%2C+Heng+Huang%2C+and+Chunyuan+Li%2E+2025%2E+Llava%2Dcritic%3A+Learning+to+evaluate+multimodal+models%2E+In+Proceedings+of+the+IEEE%2FCVF+Conferen)
44. Zhijian Xu, Yilun Zhao, Manasi Patwardhan, Lovekesh Vig, and Arman Cohan. 2025. [Can llms identify critical limitations within scientific research? A systematic evaluation on AI research papers](https://md.research.elamur.ai/free/en/papers/xu-2025-can-llms-identify-critical-limitations/). In Proceedings of the Annual Meeting of the Association for Computational Linguistic, pages 20652–20706.
45. Cheng Yang, Chufan Shi, Yaxin Liu, Bo Shui, Junjie Wang, Mohan Jing, Linran Xu, Xinyu Zhu, Siheng Li, Yuxiang Zhang, Gongye Liu, Xiaomei Nie, Deng Cai, and Yujiu Yang. 2025. Chartmimic: Evaluating lmm’s cross-modal reasoning capability via chart-to-code generation. In Proceedings of the Internationa… [find](https://scholar.google.com/scholar?q=Cheng+Yang%2C+Chufan+Shi%2C+Yaxin+Liu%2C+Bo+Shui%2C+Junjie+Wang%2C+Mohan+Jing%2C+Linran+Xu%2C+Xinyu+Zhu%2C+Siheng+Li%2C+Yuxiang+Zhang%2C+Gongye+Liu%2C+Xiaomei+Nie%2C+Deng+Cai%2C+and+Yujiu+Yang%2E+2025%2E+Chartmimic%3A+Evaluating+lmm)
46. Jiabo Ye, Haiyang Xu, Haowei Liu, Anwen Hu, Ming Yan, Qi Qian, Ji Zhang, Fei Huang, and Jingren Zhou. 2025. [mplug-owl3: Towards long image-sequence understanding in multi-modal large language models](https://md.research.elamur.ai/free/en/papers/ye-2024-mplug-owl3-towards-long-image/).
47. Jerrold H. Zar. 2005. Spearman rank correlation. Encyclopedia of biostatistics, 7. [find](https://scholar.google.com/scholar?q=Jerrold+H%2E+Zar%2E+2005%2E+Spearman+rank+correlation%2E+Encyclopedia+of+biostatistics%2C+7%2E)
48. Leixin Zhang, Steffen Eger, Yinjie Cheng, Weihe Zhai, Jonas Belouadi, Fahimeh Moafian, and Zhixue Zhao. 2025a. Scimage: How good are multimodal large language models at scientific text-to-image generation? In Proceedings of the International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Leixin+Zhang%2C+Steffen+Eger%2C+Yinjie+Cheng%2C+Weihe+Zhai%2C+Jonas+Belouadi%2C+Fahimeh+Moafian%2C+and+Zhixue+Zhao%2E+2025a%2E+Scimage%3A+How+good+are+multimodal+large+language+models+at+scientific+text%2Dto%2Dimage+genera)
49. Zicheng Zhang, Tengchuan Kou, Shushi Wang, Chunyi Li, Wei Sun, Wei Wang, Xiaoyu Li, Zongyu Wang, Xuezhi Cao, Xiongkuo Min, Xiaohong Liu, and Guangtao Zhai. 2025b. Q-eval-100k: Evaluating visual quality and alignment level for text-to-vision content. In Proceedings of the IEEE/CVF Conference on Compu… [find](https://scholar.google.com/scholar?q=Zicheng+Zhang%2C+Tengchuan+Kou%2C+Shushi+Wang%2C+Chunyi+Li%2C+Wei+Sun%2C+Wei+Wang%2C+Xiaoyu+Li%2C+Zongyu+Wang%2C+Xuezhi+Cao%2C+Xiongkuo+Min%2C+Xiaohong+Liu%2C+and+Guangtao+Zhai%2E+2025b%2E+Q%2Deval%2D100k%3A+Evaluating+visual+qualit)
50. Yilun Zhao, Weiyuan Chen, Zhijian Xu, Manasi Patwardhan, Chengye Wang, Yixin Liu, Lovekesh Vig, and Arman Cohan. 2025. [Abgen: Evaluating large language models in ablation study design and evaluation for scientific research](https://md.research.elamur.ai/free/en/papers/zhao-2025-abgen-evaluating-large-language-models/). In Proceedings of the Annual Meeting of the Association for Computational Li…
51. Bocheng Zou, Mu Cai, Jianrui Zhang, and Yong Jae Lee. 2024. Vgbench: Evaluating large language models on vector graphics understanding and generation. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 3647–3659. [find](https://scholar.google.com/scholar?q=Bocheng+Zou%2C+Mu+Cai%2C+Jianrui+Zhang%2C+and+Yong+Jae+Lee%2E+2024%2E+Vgbench%3A+Evaluating+large+language+models+on+vector+graphics+understanding+and+generation%2E+In+Proceedings+of+the+Conference+on+Empirical+Met)
52. Yuxin Wu, Alexander Kirillov, Francisco Massa, Wan-Yen Lo, and Ross Girshick. 2019. Detectron2. [https://github.com/](https://github.com/) facebookresearch/detectron2. ing of pixels), and vector graphics (consisting of basic geometric elements such as points, lines, and curves) (Yang et al., 2025). For simplification, we… [↗](https://github.com/)
53. The appendix is structured as follows: • Comparison between real captions and synthetic prompts in Section A. • Details of related work in Section B. B.1 – Scientific image datasets in Section B.1. – Evaluation metrics in Section B.2. • Details of training data in Section C. • More details of HE4AFG… [find](https://scholar.google.com/scholar?q=The+appendix+is+structured+as+follows%3A+%E2%80%A2+Comparison+between+real+captions+and+synthetic+prompts+in+Section+A%2E+%E2%80%A2+Details+of+related+work+in+Section+B%2E+B%2E1+%E2%80%93+Scientific+image+datasets+in+Section+B%2E1%2E+%E2%80%93+)

*Updated 2 November 2026*