LumiGen: An LVLM-Enhanced Iterative Framework for Fine-Grained Text-to-Image Generation

Nicholas Evans; Xiangyu Zhou; Xiaoqi Dong; Yujia Lin

REVIEW 3 major objections 5 minor 33 references

LumiGen: An LVLM-Enhanced Iterative Framework for Fine-Grained Text-to-Image Generation

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read LumiGen wraps a base diffusion text-to-image model with an LVLM that augments the prompt and then visually critiques each draft, reporting a 3.08 average human score on LongBench-T2I that beats FLUX1-dev, Omnigen, and Janus-pro-7B.

desk verdict A plausible wrapper idea with clear writing, but the results are self-admittedly fictitious and the refinement mechanism is never implemented, so it has no empirical or algorithmic weight. read the letter →

arxiv 2508.04732 v1 pith:MZZTVC5L submitted 2025-08-05 cs.LG cs.GR

Xiaoqi Dong , Xiangyu Zhou , Nicholas Evans , Yujia Lin This is my paper

classification cs.LGcs.GR

keywords text-to-imagegenerationdiffusionmodelsvision-languageiterativerefinementvisualcriticpromptaugmentationLongBench-T2Ifine-grainedcontrol

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

LumiGen tries to fix the places where text-to-image diffusion models fail: rendering legible text inside images, producing natural human poses, and keeping complex scenes coherent. Its bet is that a vision-language model, used twice—once to rewrite the prompt into a richer structured instruction, and then as a visual critic that inspects each draft image and issues correction instructions—can steer a base diffusion model toward those fine-grained goals. The paper reports this closed loop outperforms prior baselines on the LongBench-T2I benchmark with an average human score of 3.08, with the largest margins in Text (2.60) and Pose (2.58). A sympathetic reader would care because the approach is model-agnostic: if true, it offers a plug-in path to better fine-grained control without retraining the heavy generation model.

What carries the argument

The load-bearing machinery is the closed-loop pairing of Eqs. (1)-(4): $f_{\text{parse}}$ (the IPPA LVLM) turns a raw prompt into $P_{\text{aug}}$; $f_{\text{critic}}$ (the IVFR LVLM) turns the triple $(P_{\text{raw}}, P_{\text{aug}}, I_k)$ into linguistic corrections $C_k$; $h_{\text{translate}}$ turns $C_k$ into low-level control signals $\Sigma_k$; and $g_{\text{refine}}$ applies $\Sigma_k$ inside the base diffusion model. The paper states that the signals 'can manifest as' modified prompts, pose skeletons, inpainting masks, or attention-map guidance, but it never specifies, implements, or evaluates $h_{\text{translate}}$. That unspecified conversion is the single joint on which the entir

What would settle it

Instrument a reproduction of LumiGen at Eq. (3): log each correction instruction $C_k$ and record what control signal $\Sigma_k$ the base diffusion model actually receives. If $\Sigma_k$ is empty, a verbatim rewrite of the prompt, or merely a fresh random seed, the IVFR loop is not doing the claimed refinement. Equivalently, compare LumiGen against a variant that re-generates the image from scratch $N$ times with reworded prompts and no visual critique: identical average scores would show the critic loop adds nothing.

Watch

Extended reading notes

Core claim

LumiGen is a model-agnostic wrapper with two modules: IPPA rewrites a raw prompt $P_{\text{raw}}$ into a richer $P_{\text{aug}}$; then IVFR has an LVLM compare $P_{\text{raw}}$, $P_{\text{aug}}$, and the current image $I_k$, produce correction instructions $C_k$, and translate them through $h_{\text{translate}}$ into control signals $\Sigma_k$ that the base model applies via $g_{\text{refine}}$ to yield $I_{k+1}$. The paper reports the loop scores 3.08 on LongBench-T2I, beating Omnigen (2.96), FLUX1-dev (2.78), and Janus-pro-7B (2.50), with the biggest gains in Text (2.60) and Pose (2.58). Ablations attribute 0.12 of the average to IPPA and 0.18 to IVFR; five iterations raise the score from

Load-bearing premise

The refinement loop only works if the unstated translation function $h_{\text{translate}}$ can convert the LVLM's natural-language correction instructions into concrete low-level control signals (masks, pose skeletons, attention guidance) that the base diffusion model can actually apply; the paper gives no implementation or test of that translation.

Editorial extensions

If this is right

Because LumiGen wraps an existing diffusion model rather than retraining it, the reported gains are additive on top of whatever the base model already does well.
The IVFR loop is specifically credited with gains in text rendering and pose expression, the two dimensions where one-pass diffusion models have been weakest.
Ablation shows both modules are needed: dropping IPPA lowers the average from 3.08 to 2.96, and dropping IVFR lowers it to 2.90.
Improvement is monotonic across refinement rounds, rising from 2.90 after initial generation to 3.08 after five iterations, with gains largely saturated by iterations three to five.

Reading between the lines

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

If $h_{\text{translate}}$ is realized with pose skeletons and inpainting masks, LumiGen is effectively a prompt-engineering wrapper over the base model's built-in conditioning mechanisms; the reported gains may come from the IPPA rewrite plus multi-pass resampling rather than from true closed-loop visual correction.
The benchmark's human evaluators may reward the longer, more detailed augmented prompts regardless of alignment; a controlled test holding prompt verbosity fixed would separate the augmentation effect from the critic effect.
The convergence at three to five iterations suggests a fixed iteration budget is sufficient, but an adaptive stopping rule based on LVLM confidence could trade compute for better quality—an extension the paper lists as future work.
The same critic loop could in principle be applied to video or 3D generation, since the correction format (linguistic directives plus masks or skeletons) is not tied to image-specific internals; the paper only mentions video/3D as future work, so this is an extrapolation.

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.

Desk Editor's Note

read the letter

Here's the short version: LumiGen is a wrapper idea—use an LVLM to parse and augment the prompt, generate an image, then have the LVLM criticize the image and feed corrections back into the diffusion model—and the paper describes it in a clear, readable way. The related work is on point and it honestly situates itself in the iterative-feedback line. But the only quantitative support is Tables 1–3, and every one of those captions says 'Scores are fictitious for demonstration purposes.' So the headline 3.08 vs. 2.96 is not a measured result; it's a placeholder. The abstract and conclusions cite it without the caveat, which is misleading.

The second problem is that the core loop is specified only at the level of equations (2)–(4). We're told C_k = f_critic(...), Sigma_k = h_translate(C_k), and I_{k+1} = g_refine(...), but h_translate and g_refine are never instantiated. How does 'make the word Eureka more distinct' become a mask, a pose skeleton, or attention guidance? No algorithm, no implementation detail, no test. Without that, the claimed Text and Pose gains cannot occur—the loop has no mechanism. Section 4.1 says they use Stable Diffusion XL and LLaVA, but fine-tuning details are absent.

There's also a minor citation slip: [6] and [33] are the same paper (Draw ALL your imagine), cited twice with different numbers.

So what's the value here? As a position piece, it's a fair summary of an idea that is already in the literature—the paper itself cites [27] and [32] for the same pattern. It adds no new benchmark, no formal result, no code, and no real evaluation. The fictitious tables are not a small flaw; they remove the empirical backbone of the entire paper.

My recommendation: this should be desk-rejected, not sent to peer review. There is nothing for a referee to check. If the authors actually build the h_translate/g_refine pipeline and run a real human evaluation on LongBench-T2I, then it becomes a legitimate (if incremental) systems paper. As it stands, it is an architecture sketch with invented numbers.

Referee Report

3 major / 5 minor

Summary. LumiGen is an iterative text-to-image framework that wraps a base diffusion model with two LVLM-driven modules: IPPA (prompt parsing/augmentation) and IVFR (visual critique and correction). The paper formalizes the pipeline in Eqs. (1)-(4) as fparse, fcritic, htranslate, and grefine, and reports human-evaluation scores on LongBench-T2I claiming an average of 3.08 versus 2.96 for Omnigen, with strengths in Text (2.60) and Pose (2.58). Ablation and iteration analyses in §4.4-4.5 attribute the gains to both modules.

Significance. The conceptual idea of using an LVLM as a planner and critic in a closed loop is timely and potentially useful, and the paper is clearly written at a high level. However, the manuscript contains no actual empirical evidence. Tables 1-3 are explicitly labeled 'fictitious for demonstration purposes,' and the translation mechanism htranslate that is essential for refinement is never specified. As submitted, the central claims are unsupported, so the significance cannot be assessed.

major comments (3)

[§4.2, Tables 1-3] The captions of all three quantitative tables state 'Scores are fictitious for demonstration purposes.' The abstract, §4.3, and §5 nonetheless report 3.08 vs. 2.96 and Text/Pose gains as measured results. Because the only quantitative support for the headline claims is explicitly fictitious, the paper's central empirical assertion is unsupported. This is not a minor caveat; it undermines the evaluation and conclusions.
[§3.3, Eqs. (2)-(4)] The refinement loop requires htranslate to convert linguistic correction instructions Ck into low-level control signals Σk (masks, pose skeletons, attention guidance), and grefine to apply them inside the base T2I model. The paper describes these functions only verbally; no algorithm, training procedure, or implementation is given. §4.1 names Stable Diffusion XL and LLaVA but does not explain how corrections are channeled into the diffusion process. Consequently the IVFR loop that is claimed to drive the Text and Pose gains is not actually specified, and the reported improvement cannot be reproduced or even traced to a concrete mechanism.
[§4.1] Even if the scores were real, the evaluation report omits essential methodology: number of prompts used, number of human evaluators, inter-rater agreement, and statistical significance. The reported differences between LumiGen (3.08) and Omnigen (2.96) are small; without significance testing or variance information, the claimed superiority is not established. Please provide full evaluation protocol details in any revision.

minor comments (5)

[References [6] and [33]] References [6] and [33] refer to the same work (Draw ALL your imagine); they should be unified to avoid duplication and citation inconsistency.
[Title] The title contains an apparent spacing artifact: 'L VLM-Enhanced' should be 'LVLM-Enhanced'.
[Table 4] The caption of Table 4 says 'Scores are from human feedback,' but the table contains qualitative strength descriptions and no numerical scores; the caption is misleading.
[§2.1, ref [12]] The citation [12] (DreamBooth) is used for Diff-Tuning and 'chain of forgetting'; this appears mismatched. Please verify that all related-work citations accurately correspond to the described methods.
[§3.4] The number of refinement iterations N is a free parameter with no sensitivity analysis or selection criterion provided; please report its value and robustness in the experimental section.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity in the derivation chain; main quantitative claim is unsupported because evaluation tables are explicitly labeled fictitious.

full rationale

The paper's derivation chain is an architectural proposal: Eq. (1) defines prompt augmentation Paug=fparse(Praw); Eqs. (2)-(4) define the refinement loop Ck=fcritic(...), Sigma_k=htranslate(Ck), and I_{k+1}=grefine(...). These are definitions/placeholders, not derivations, and no predicted quantity is shown to equal an input by construction. The claimed result (avg 3.08) is not computed in the text; it appears only in Table 1, whose caption states 'Scores are fictitious for demonstration purposes' (same for Tables 2 and 3). That is a missing-evidence/data-integrity problem, not a circularity: the number is not a fitted parameter renamed as a prediction, it is a placeholder presented as a result. The paper does cite prior work by the same author group ([5], [6]/[33]), and the benchmark is co-authored by the second author, but these citations are not used to force a conclusion by an unverified theorem; the empirical conclusion is unsupported directly by the fictitious tables. Therefore no step reduces to its own input, and the circularity score is 0. The severe unsupportedness of the headline result should be handled as a correctness/evidence issue.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central claim rests on: an assumed LVLM critic that gives accurate corrections, an unspecified translator htranslate from text to control signals, an assumed base diffusion model that can consume those signals, and a real human evaluation that the tables themselves contradict. No experiment or code backs these assumptions.

free parameters (1)

Number of refinement iterations N = not specified
Section 3.4 defines the loop as running for a predefined N, but gives no value, schedule, or stopping criterion; the claimed convergence in Section 4.5 depends on N.

assumptions (5)

domain assumption A pretrained LVLM can produce reliable visual criticism that identifies real discrepancies between image and prompt.
Section 3.3 defines fcritic but gives no evidence that such criticism is accurate or actionable; false criticism would break the refinement loop.
ad hoc to paper Linguistic correction instructions can be translated into low-level control signals via htranslate.
Eq. (3) defines htranslate, but no implementation, training, or concrete examples are provided anywhere in the paper.
domain assumption The underlying T2I model can consume control signals, such as masks or pose skeletons, and perform local refinement in place.
Eq. (4) depends on grefine; the paper assumes a fine-tuned diffusion model supports such interventions, which is not demonstrated.
domain assumption A LongBench-T2I human evaluation was actually conducted with professional evaluators.
Tables 1, 2, and 3 state that scores are fictitious, so this assumption is contradicted by the manuscript itself.
domain assumption The augmented prompt Paug improves rather than distorts user intent.
Section 3.2 assumes IPPA parsing adds useful detail; no user study or metric confirms that the expanded prompts preserve the user's intent.

how reviews work

0 comments

Cite this review

Pith. "Pith review of LumiGen: An LVLM-Enhanced Iterative Framework for Fine-Grained Text-to-Image Generation." pith.science (2026). https://pith.science/paper/MZZTVC5L

@misc{pith2026250804732,
  author       = {Pith},
  title        = {Pith review of: LumiGen: An LVLM-Enhanced Iterative Framework for Fine-Grained Text-to-Image Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MZZTVC5L}},
  note         = {Machine review of arXiv:2508.04732}
}

read the original abstract

Text-to-Image (T2I) generation has made significant advancements with diffusion models, yet challenges persist in handling complex instructions, ensuring fine-grained content control, and maintaining deep semantic consistency. Existing T2I models often struggle with tasks like accurate text rendering, precise pose generation, or intricate compositional coherence. Concurrently, Vision-Language Models (LVLMs) have demonstrated powerful capabilities in cross-modal understanding and instruction following. We propose LumiGen, a novel LVLM-enhanced iterative framework designed to elevate T2I model performance, particularly in areas requiring fine-grained control, through a closed-loop, LVLM-driven feedback mechanism. LumiGen comprises an Intelligent Prompt Parsing & Augmentation (IPPA) module for proactive prompt enhancement and an Iterative Visual Feedback & Refinement (IVFR) module, which acts as a "visual critic" to iteratively correct and optimize generated images. Evaluated on the challenging LongBench-T2I Benchmark, LumiGen achieves a superior average score of 3.08, outperforming state-of-the-art baselines. Notably, our framework demonstrates significant improvements in critical dimensions such as text rendering and pose expression, validating the effectiveness of LVLM integration for more controllable and higher-quality image generation.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 16 canonical work pages

[6]

arXiv preprint arXiv:2505.24787 (2025)

Zhou, Y., Yuan, J., Wang, Q.: Draw all your imagine: A holistic benchmark and agent framework for complex instruction-based image generation. arXiv preprint arXiv:2505.24787 (2025)

arXiv 2025
[33]

CoRR (2025)

Zhou, Y., Yuan, J., Wang, Q.: Draw ALL your imagine: A holistic benchmark and agent framework for complex instruction-based image generation. CoRR (2025). https://doi.org/10.48550/ARXIV.2505.24787

arXiv doi:10.48550/arxiv.2505.24787 2025
[27]

Singh, J., Zheng, L.: Divide, evaluate, and refine: Evaluating and improving text- to-image alignment with iterative VQA feedback. In: Advances in Neural Informa- tion Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023 (2023)

work page 2023
[32]

CoRR (2024)

Duan, P., Chen, C., Hartmann, B., Li, Y.: Visual prompting with iterative refinement for design critique generation. CoRR (2024). https://doi.org/10.48550/ARXIV.2412.16829

arXiv doi:10.48550/arxiv.2412.16829 2024
[1]

IEEE Trans

He, C., Shen, Y., Fang, C., Xiao, F., Tang, L., Zhang, Y., Zuo, W., Guo, Z., Li, X.: Diffusion models in low-level vision: A survey. IEEE Trans. Pattern Anal. Mach. Intell. pp. 4630–4651 (2025). https://doi.org/10.1109/TPAMI.2025.3545047

work page arXiv 2025
[2]

Frontiers Comput

Zhu, B., Zhang, H.: Debiasing vision-language models for vision tasks: a sur- vey. Frontiers Comput. Sci. p. 191321 (2025). https://doi.org/10.1007/S11704-024- 40051-3

doi:10.1007/s11704-024- 2025
[3]

In: Findings of the Association for Computational Linguistics, ACL 2024, Bangkok, Thailand and virtual meeting, August 11-16, 2024

Zhou, Y., Li, X., Wang, Q., Shen, J.: Visual in-context learning for large vision- language models. In: Findings of the Association for Computational Linguistics, ACL 2024, Bangkok, Thailand and virtual meeting, August 11-16, 2024. pp. 15890– 15902. Association for Computational Linguistics (2024)

work page 2024
[4]

arXiv preprint arXiv:2402.07398 (2024)

Zhu, D., Tang, X., Han, W., Lu, J., Zhao, Y., Xing, G., Wang, J., Yin, D.: Vis- linginstruct: Elevating zero-shot learning in multi-modal language models with autonomous instruction optimization. arXiv preprint arXiv:2402.07398 (2024)

arXiv 2024

Show all 33 references

[5]

arXiv preprint arXiv:2501.01377 (2025)

Zhou, Y., Song, L., Shen, J.: Improving medical large vision-language models with abnormal-aware feedback. arXiv preprint arXiv:2501.01377 (2025)

2025 arXiv
[7]

In: IEEE/CVF Win- ter Conference on Applications of Computer Vision, WACV 2025 - Workshops, Tucson, AZ, USA, February 28 - March 4, 2025

Ciamarra, A., Caldelli, R., Bimbo, A.D.: On the generalisation capability of lo- cal surface frames in detecting diffusion-based facial images. In: IEEE/CVF Win- ter Conference on Applications of Computer Vision, WACV 2025 - Workshops, Tucson, AZ, USA, February 28 - March 4, 2...

2025
[8]

In: IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025, Nashville, TN, USA, June 11-15, 2025

Xiao, S., Wang, Y., Zhou, J., Yuan, H., Xing, X., Yan, R., Li, C., Wang, S., Huang, T., Liu, Z.: Omnigen: Unified image generation. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025, Nashville, TN, USA, June 11-15, 2025. pp. 13294–13304. Computer Vi...

2025
[9]

CoRR (2025)

Chen, X., Wu, Z., Liu, X., Pan, Z., Liu, W., Xie, Z., Yu, X., Ruan, C.: Janus-pro: Unified multimodal understanding and generation with data and model scaling. CoRR (2025). https://doi.org/10.48550/ARXIV.2501.17811

2025 doi arXiv
[10]

Huang, K., Sun, K., Xie, E., Li, Z., Liu, X.: T2i-compbench: A comprehensive benchmark for open-world compositional text-to-image generation. In: Advances in Neural Information Processing Systems 36: Annual Conference on Neural Infor- mation Processing Systems 2023, NeurIPS 20...

2023
[11]

arXiv preprint arXiv:2506.12830 (2025)

Wang, C., Zhou, Y., Wang, Q., Wang, Z., Zhang, K.: Complexbench-edit: Bench- marking complex instruction-driven image editing via compositional dependencies. arXiv preprint arXiv:2506.12830 (2025)

2025 arXiv
[12]

In: IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, Vancouver, BC, Canada, June 17-24, 2023

Ruiz, N., Li, Y., Jampani, V., Pritch, Y., Rubinstein, M., Aberman, K.: Dream- booth: Fine tuning text-to-image diffusion models for subject-driven generation. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023, Vancouver, BC, Canada, June 17-24, 202...

2023
[13]

CoRR (2023)

Zhang, X., Zhao, W., Lu, X., Chien, J.: Text2layer: Layered image generation using latent diffusion model. CoRR (2023). https://doi.org/10.48550/ARXIV.2307.09781

2023 doi arXiv
[14]

In: IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022

Zhou, Y., Zhang, R., Chen, C., Li, C., Tensmeyer, C., Yu, T., Gu, J., Xu, J., Sun, T.: Towards language-free training for text-to-image generation. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022. pp. 17886–...

2022
[15]

Yu, J., Xu, Y., Koh, J.Y., Luong, T., Baid, G., Wang, Z., Vasudevan, V., Ku, A., Yang, Y., Ayan, B.K., Hutchinson, B., Han, W., Parekh, Z., Li, X., Zhang, H., Baldridge, J., Wu, Y.: Scaling autoregressive models for content-rich text-to-image generation. Trans. Mach. Learn. Re...

2022
[16]

In: Pattern Recognition - 27th International Conference, ICPR 2024, Kolkata, India, December 1-5, 2024, Proceedings, Part VI

Gu,S.,Su,J.,Duan,Y.,Chen,X.,Luo,J.,Zhao,H.:Text2street:Controllabletext- to-image generation for street views. In: Pattern Recognition - 27th International Conference, ICPR 2024, Kolkata, India, December 1-5, 2024, Proceedings, Part VI. pp. 130–145. Springer (2024). https://do...

2024 doi
[17]

In: IEEE Conference on Computer Vi- sion and Pattern Recognition, CVPR 2019, Long Beach, CA, USA, June 16-20, 2019

Yin, G., Liu, B., Sheng, L., Yu, N., Wang, X., Shao, J.: Semantics disen- tangling for text-to-image generation. In: IEEE Conference on Computer Vi- sion and Pattern Recognition, CVPR 2019, Long Beach, CA, USA, June 16-20, 2019. pp. 2327–2336. Computer Vision Foundation / IEEE...

2019
[18]

IEEE Trans

Feng, Y., Wang, X., Wong, K., Wang, S., Lu, Y., Zhu, M., Wang, B., Chen, W.: Promptmagician: Interactive prompt engineering for text- to-image creation. IEEE Trans. Vis. Comput. Graph. pp. 295–305 (2024). https://doi.org/10.1109/TVCG.2023.3327168

2024
[19]

IEEE Trans

Xu, P., Shao, W., Zhang, K., Gao, P., Liu, S., Lei, M., Meng, F., Huang, S., Qiao, Y., Luo, P.: Lvlm-ehub: A comprehensive evaluation benchmark for large vision- language models. IEEE Trans. Pattern Anal. Mach. Intell. pp. 1877–1893 (2025). https://doi.org/10.1109/TPAMI.2024.3...

2025
[20]

CoRR (2024)

Liang, C.X., Tian, P., Yin, C.H., Yua, Y., An-Hou, W., Ming, L., Wang, T., Bi, Z., Liu, M.: A comprehensive survey and guide to multimodal large language models in vision-language tasks. CoRR (2024). https://doi.org/10.48550/ARXIV.2411.06284

2024 doi
[21]

In: The Thirteenth International Conference on Learning Representations (2025)

Zhou, Y., Shen, J., Cheng, Y.: Weak to strong generalization for large language models with multi-capabilities. In: The Thirteenth International Conference on Learning Representations (2025)

2025
[22]

CoRR (2024)

Ghosh, A., Acharya, A., Saha, S., Jain, V., Chadha, A.: Exploring the frontier of vision-language models: A survey of current methodologies and future directions. CoRR (2024). https://doi.org/10.48550/ARXIV.2404.07214

2024 doi
[23]

Dai, W., Li, J., Li, D., Tiong, A.M.H., Zhao, J., Wang, W., Li, B., Fung, P., Hoi, S.C.H.: Instructblip: Towards general-purpose vision-language models with instruction tuning. In: Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Pr...

2023
[24]

arXiv preprint arXiv:2210.03963 (2022)

Zhu, D., Mao, Z., Lu, J., Zhao, R., Tan, F.: Sda: simple discrete augmentation for contrastive sentence representation learning. arXiv preprint arXiv:2210.03963 (2022)

2022 arXiv
[25]

In: Findings of the Association for Computational Linguistics: EACL 2023

Zhou, Y., Long, G.: Style-aware contrastive learning for multi-style image caption- ing. In: Findings of the Association for Computational Linguistics: EACL 2023. pp. 2257–2267 (2023)

2023
[26]

In: ICASSP 2021-2021 IEEE International Conference onAcoustics,SpeechandSignalProcessing(ICASSP).pp.7598–7602.IEEE(2021)

Zhou, Y., Tao, W., Zhang, W.: Triple sequence generative adversarial nets for unsu- pervised image captioning. In: ICASSP 2021-2021 IEEE International Conference onAcoustics,SpeechandSignalProcessing(ICASSP).pp.7598–7602.IEEE(2021)

2021
[28]

arXiv preprint arXiv:2506.19835 (2025)

Zhou, Y., Song, L., Shen, J.: Mam: Modular multi-agent framework for multi-modal medical diagnosis via role-specialized collaboration. arXiv preprint arXiv:2506.19835 (2025)

2025 arXiv
[29]

arXiv preprint arXiv:2501.12432 (2025)

Zhu, D., Shi, W., Shi, Z., Ren, Z., Wang, S., Yan, L., Yin, D.: Divide-then- aggregate: An efficient tool learning method via parallel tool invocation. arXiv preprint arXiv:2501.12432 (2025)

2025 arXiv
[30]

In: Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10, 2023

You, H., Sun, R., Wang, Z., Chen, L., Wang, G., Ayyubi, H.A., Chang, K., Chang, S.: Idealgpt: Iteratively decomposing vision and language reasoning via large lan- guage models. In: Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10,...

2023 doi
[31]

CoRR (2024)

Chen, Z., Yang, L., Chen, S., Chen, Z., Liang, J., Li, X.: Revis- iting prompt pretraining of vision-language models. CoRR (2024). https://doi.org/10.48550/ARXIV.2409.06166

2024 doi arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [6]

arXiv preprint arXiv:2505.24787 (2025)

Zhou, Y., Yuan, J., Wang, Q.: Draw all your imagine: A holistic benchmark and agent framework for complex instruction-based image generation. arXiv preprint arXiv:2505.24787 (2025)

arXiv 2025

[2] [33]

CoRR (2025)

Zhou, Y., Yuan, J., Wang, Q.: Draw ALL your imagine: A holistic benchmark and agent framework for complex instruction-based image generation. CoRR (2025). https://doi.org/10.48550/ARXIV.2505.24787

arXiv doi:10.48550/arxiv.2505.24787 2025

[3] [27]

Singh, J., Zheng, L.: Divide, evaluate, and refine: Evaluating and improving text- to-image alignment with iterative VQA feedback. In: Advances in Neural Informa- tion Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023 (2023)

work page 2023

[4] [32]

CoRR (2024)

Duan, P., Chen, C., Hartmann, B., Li, Y.: Visual prompting with iterative refinement for design critique generation. CoRR (2024). https://doi.org/10.48550/ARXIV.2412.16829

arXiv doi:10.48550/arxiv.2412.16829 2024

[5] [1]

IEEE Trans

He, C., Shen, Y., Fang, C., Xiao, F., Tang, L., Zhang, Y., Zuo, W., Guo, Z., Li, X.: Diffusion models in low-level vision: A survey. IEEE Trans. Pattern Anal. Mach. Intell. pp. 4630–4651 (2025). https://doi.org/10.1109/TPAMI.2025.3545047

work page arXiv 2025

[6] [2]

Frontiers Comput

Zhu, B., Zhang, H.: Debiasing vision-language models for vision tasks: a sur- vey. Frontiers Comput. Sci. p. 191321 (2025). https://doi.org/10.1007/S11704-024- 40051-3

doi:10.1007/s11704-024- 2025

[7] [3]

In: Findings of the Association for Computational Linguistics, ACL 2024, Bangkok, Thailand and virtual meeting, August 11-16, 2024

Zhou, Y., Li, X., Wang, Q., Shen, J.: Visual in-context learning for large vision- language models. In: Findings of the Association for Computational Linguistics, ACL 2024, Bangkok, Thailand and virtual meeting, August 11-16, 2024. pp. 15890– 15902. Association for Computational Linguistics (2024)

work page 2024

[8] [4]

arXiv preprint arXiv:2402.07398 (2024)

Zhu, D., Tang, X., Han, W., Lu, J., Zhao, Y., Xing, G., Wang, J., Yin, D.: Vis- linginstruct: Elevating zero-shot learning in multi-modal language models with autonomous instruction optimization. arXiv preprint arXiv:2402.07398 (2024)

arXiv 2024