Pith. sign in

REVIEW 3 major objections 4 minor 2 cited by

A training-free pipeline aligns text-to-image generation with a user's preferences in real time, using MLLM-extracted keywords and global-local attention guidance from a single reference image.

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 ↗

An MLLM-driven, training-free pipeline extracts preference keywords from a reference image and modulates diffusion cross-attention at global and regional levels for instant, multi-round preference-aligned image generation.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection A plausible training-free preference-alignment framework, undermined by its own supplementary time table showing ~180s/image—the 'real-time' claim is internally contradicted. the 3 major comments →

arxiv 2508.17718 v1 pith:ZJI67MCY submitted 2025-08-25 cs.CV cs.AI

Instant Preference Alignment for Text-to-Image Diffusion Models

Yang Li , Songlin Yang , Xiaoxuan Han , Wei Wang , Jing Dong , Yueming Lyu , Ziyu Xue This is my paper
classification cs.CV cs.AI
keywords text-to-image generationpreference alignmenttraining-free controlmultimodal large language modelsprompt enrichmentcross-attention modulationdiffusion modelsinteractive refinement
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 reading

This paper tries to establish that preference-aligned text-to-image generation can be done instantly and without fine-tuning: a user hands over a reference image they like, and the model generates new images that carry over the same artistic style, emotional tone, themes, and visual elements. The key move is to separate the problem into understanding preferences and using them to guide generation. Understanding is handled by a multimodal large language model that turns a reference image into structured keywords and an enriched prompt; guidance is handled at inference time by adding a projected preference embedding to the prompt embeddings and by modulating cross-attention region-wise. If this works as claimed, users and dialog systems can steer image generation in real time and iteratively, without retraining per preference or per user.

Core claim

The central claim is that the full preference-alignment loop can be moved out of training. Given a reference image, an MLLM extracts keywords in four preference categories—artistic style, emotion/atmosphere, theme, visual elements—and rewrites the base prompt into a complex prompt, per-entity sub-prompts, and a background prompt. Generation then uses two training-free controls: global preference guidance, which orthogonally projects preference keywords into each prompt embedding so the output is shifted toward reference-image attributes without losing the original prompt's meaning; and local cross-attention modulation, which renders each entity within its planned bounding box by replacing ba

What carries the argument

The machine is a two-stage inference-time pipeline. Preference understanding is carried by a fixed four-category keyword taxonomy (artistic style, emotional/atmospheric, thematic, visual elements) plus chain-of-thought MLLM prompting that extracts keywords from a reference image and enriches the base prompt. Preference-guided generation is carried by two mechanisms: global preference guidance, P_yi = T_text(y_i) + α·Proj⊥(T_text(y_key)), which injects preference keywords into text embedding space while subtracting the component already present in the prompt; and local cross-attention modulation, which computes region-specific cross-attention latents from per-entity sub-prompts and writes the

Load-bearing premise

The whole pipeline rests on the premise that a user's preference is adequately captured by MLLM-generated keywords organized into four fixed categories extracted from a single reference image—and that the evaluation's category menu does not conceal preferences outside that taxonomy.

What would settle it

Run a preference-elicitation study with fully open-ended user responses (no category menu) and check whether the MLLM keywords and generated images align with the stated reasons; if a substantial fraction of preferences fall outside the four categories or the keywords miss them, the coverage claim is falsified.

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

If this is right

  • A single reference image, not a per-user dataset, becomes the preference specification, so preference changes take effect in one generation step.
  • The framework can serve as an interactive image editor: users refine keywords, prompts, entities, or boxes across rounds, and each round amounts to inference-time control.
  • Components are swappable: the paper shows the same pipeline with different MLLMs and different diffusion backbones, so gains transfer rather than being locked to one model.
  • Because guidance lives in text-embedding and attention space, it sidesteps RLHF/DPO-style retraining and coarse binary feedback.
  • Prompt fidelity and preference alignment are treated as separable axes: the orthogonal projection preserves base-prompt meaning while preference strength is controlled by α.

Where Pith is reading between the lines

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

  • Reading beyond the paper, the ceiling of this framework likely tracks the MLLM's ability to see and verbalize a preference; replacing the MLLM with a stronger one should raise alignment without changing the guidance equations.
  • The four-category taxonomy is probably a pragmatic simplification; a natural stress test is free-form preference elicitation, since the paper's own coverage table gives users only these categories to choose from, making 'Others' at 4.3% partly an artifact of the question format.
  • The orthogonal-projection trick could be reused as a generic semantic-preservation module for any conditioning signal that lives in text space, independent of image preferences.
  • A cleaner controlled comparison—feeding the same enriched prompt to a plain rewrite baseline versus the full pipeline—would isolate how much of the gain comes from global keyword guidance versus local cross-attention modulation.
Share X Bluesky LinkedIn Reddit HN

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 proposes a training-free framework for preference-aligned text-to-image generation. Given a reference image and a base prompt, an MLLM extracts preference keywords in four fixed categories, enriches the prompt with preference-informed entities, plans bounding boxes, and then performs inference with global keyword-embedding projection (Eq. 6) and local region-wise cross-attention modulation (Eqs. 7-10). The authors report qualitative and quantitative comparisons against NP, Fabric, MGIE, Prompt Rewrite, Viper, and InstantStyle, along with ablations, multi-round interaction examples, and generalization across MLLMs and diffusion backbones. The central claim is that this is an 'instant/real-time' preference alignment method that outperforms prior approaches in both metrics and user studies.

Significance. If the central claims are upheld, the framework would offer a flexible training-free alternative to preference fine-tuning and to Viper-style personalization, with a clean global/local decomposition and a single-image preference interface. The paper is well motivated and provides extensive qualitative evidence, module ablations, and multi-round interaction examples. However, the strongest differentiators—'instant', 'real-time', and quantitative superiority—are currently not supported by the paper's own measurements and reporting. The execution-time table (Supp. Table S5) directly contradicts the headline claim, and several metrics are reported in ways that prevent independent verification. The underlying architecture remains plausible and potentially valuable, but the present manuscript needs substantial revision before the central claims can be accepted.

major comments (3)
  1. [Supp. Sec. B.2 / Table S5 vs. Abstract, Sec. 1] Table S5 reports our method's execution time as ~180 s per image, compared with ~7 s for NP and ~10 s for InstantStyle, and on par with Viper/Fabric (>180 s). This directly contradicts the abstract's 'real-time and training-free' claim and Contribution 1 ('instant preference alignment framework for real-time T2I generation'). The supplementary text even claims the method avoids 'the substantial time overhead associated with methods like Viper and Fabric,' which its own table disproves. Please provide a wall-clock breakdown (MLLM inference vs. diffusion steps vs. attention composition) and either revise the real-time/instant claims or reduce the actual latency; as written, the main differentiator over prior work is falsified by the paper's own measurement.
  2. [Sec. 4.1, Table 1 and Table 2] ImageReward is reported as dimensionless percentages (e.g., 25.4% in Table 1 and 80.2% in Table 2) without any stated normalization or win-rate protocol; ImageReward is normally a scalar reward, so the units and the text 'text corresponds to the base prompt and the extracted keywords' do not explain these numbers. In the same table, 'Style Loss' is placed under 'Metrics↑', yet a loss is conventionally lower-is-better, and Sec. 4.2 says InstantStyle (0.712, the maximum in that column) 'excels in Style Loss' while Ours is 0.663. The direction of this metric must be clarified, or the quantitative superiority claim cannot be evaluated.
  3. [Sec. 4.1, Table 3; Sec. 4.3; Limitations paragraph] The completeness of the four-category preference taxonomy is validated by asking users to choose reasons from the same four categories (plus 'Others'); with 4.3% choosing 'Others', the 95.7% coverage is partly a construction. The method and its evaluation rely on MLLMs (Qwen-VL-72B for extraction; GPT-4o as judge), so shared MLLM bias is a real risk, and the human study (20 participants, no significance testing) is not sufficient to rule it out. The paper's own Limitations paragraph concedes that 'certain minor preferences may not be fully captured,' which further tempers the 'broader and more fine-grained coverage' claim. Please add an open-ended preference-elicitation study and independent human judgments with error bars.
minor comments (4)
  1. [Eq. (7)] Typo: 'generate generate corresponding' should be 'generate corresponding'.
  2. [Sec. 4.2] The text says 'This distribution of these selections (Fig. 3, first row)' but the referred result is in Table 3, not Fig. 3.
  3. [Sec. 4.1] ImageReward is cited as (Gal et al. 2022); the reference list attributes ImageReward to Xu et al. (2023). Please correct the citation.
  4. [Limitations and Future Work] The text says 'as shown in failure cases in Supp. C,' but Supp. C does not contain a failure-case section. Either provide the promised failure cases or correct the reference.

Circularity Check

2 steps flagged

Two evaluation anchors are self-referential: ImageReward is scored against the method's own extracted keywords, and the four-category preference taxonomy is 'validated' by a forced-choice instrument that only offers those categories.

specific steps
  1. fitted input called prediction [Sec. 4.1 Evaluation Metrics; Eq. (4)-(6)]
    "CLIP Score ... and ImageReward ... are adopted for assessing the text-image alignment, where the text corresponds to the base prompt and the extracted keywords, respectively."

    The 'extracted keywords' are produced by Eq. (4) from the reference image and then used as conditioning in Eq. (5) and Eq. (6): Pyi = Ttext(yi) + α·Proj⊥(Ttext(ykey)). ImageReward is then computed with exactly the same ykey as the target text. Thus the metric measures whether the output matches the method's own conditioning signal, not whether those keywords correspond to independently verified user preference. High ImageReward is partially built in because the generator is explicitly steered toward ykey. The only external check on ykey (Table 3) is itself a forced-choice instrument, so the chain is not independently anchored.

  2. self definitional [Sec. 4.1 Preference Understanding / Table 3; Sec. 1]
    "Participants are presented with a set of images, asked to select their preferred one, and then specify their preference reasons from: artistic style, emotional resonance, thematic, visual elements, and other reasons. This distribution of these selections (Fig. 3, first row) confirms the completeness of our categories. Notably, with only 4.3% of users choosing 'Others'."

    The paper's claim that its four categories 'cover the vast majority of user preferences' (Sec. 1) is validated by an instrument in which participants can only choose among the four predefined categories plus an 'Others' residual. The 4.3% 'Others' rate is therefore partly a construction of the forced-choice format: any preference outside the taxonomy can only be expressed through the residual bucket, and that bucket's small size is then used to declare the taxonomy complete. This is self-definitional support, not an independent measurement of coverage.

full rationale

The core generation pipeline (Eqs. 5-10) is a concrete, training-free mechanism applied to SDXL, and the comparisons against Viper, InstantStyle, Fabric, MGIE, and NP are externally benchmarked, including a human pairwise study. So the paper is not wholly circular. However, two load-bearing validation steps reduce to the paper's own constructs. First, the ImageReward 'keyword alignment' metric uses the MLLM-extracted keywords that are themselves part of the generation conditioning (Eqs. 4-6) as the evaluation text, so a high score partly reflects the method chasing its own conditioning signal rather than independently verified user preference. Second, the claimed completeness of the four-category preference taxonomy is validated by a forced-choice instrument whose only escape is an 'Others' bucket, making the 95.7% coverage (4.3% Others) partly tautological. The shared MLLM-as-judge design (Qwen-VL for extraction, GPT-4o for evaluation) is a bias risk but not a formal reduction, and the Supp. Table S5 ~180s timing contradicts the 'instant' headline but is a correctness/consistency issue, not a circular-derivation issue. Hyperparameter choices (α, λ) are tuned on the reported benchmarks, which weakens the generality of the numbers but does not by itself make the derivation circular. Because two evaluation anchors are self-referential, the overall circularity score is 6.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

The framework rests on four domain assumptions about MLLM reliability and text-embedding geometry, plus two tuned hyper-parameters. It introduces no new physical or conceptual entity.

free parameters (2)
  • alpha (global preference guidance strength) = 0.7
    Chosen by ablation (Fig. S12) on the same test set used for the reported results; no held-out validation.
  • lambda (complex-prompt vs composed-latent blend weight) = 0.2
    Chosen by ablation (Fig. S13) on the same benchmark images; affects entity blending in Eq. 10.
axioms (4)
  • domain assumption MLLM can distill user preference from a single reference image into keyword lists (Eq. 4)
    The whole preference-understanding stage depends on Qwen-VL correctly verbalizing what a user likes; no mechanism outside MLLM output validates this.
  • domain assumption Four preference categories (Artistic Style, Emotional, Thematic, Visual Elements) plus 'Others' cover almost all user preferences
    Section 3.2 and Table 3; validated only by a forced-choice user study where participants pick from these same categories.
  • domain assumption Text embedding space is Euclidean enough that adding an orthogonalized preference embedding preserves prompt semantics (Eq. 6)
    The global guidance method assumes Proj_perp(T_text(y_key)) removes prompt-related content while retaining preference; no proof, only qualitative ablations.
  • domain assumption Masked latent blending in cross-attention (Eq. 8) yields coherent multi-entity images without training
    Regional cross-attention modulation assumes the composed latent features blend seamlessly; failures shown only qualitatively in Supp. Fig. S11.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Instant Preference Alignment for Text-to-Image Diffusion Models." pith.science (2026). https://pith.science/paper/ZJI67MCY

@misc{pith2026250817718,
  author       = {Pith},
  title        = {Pith review of: Instant Preference Alignment for Text-to-Image Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZJI67MCY}},
  note         = {Machine review of arXiv:2508.17718}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Text-to-image (T2I) generation has greatly enhanced creative expression, yet achieving preference-aligned generation in a real-time and training-free manner remains challenging. Previous methods often rely on static, pre-collected preferences or fine-tuning, limiting adaptability to evolving and nuanced user intents. In this paper, we highlight the need for instant preference-aligned T2I generation and propose a training-free framework grounded in multimodal large language model (MLLM) priors. Our framework decouples the task into two components: preference understanding and preference-guided generation. For preference understanding, we leverage MLLMs to automatically extract global preference signals from a reference image and enrich a given prompt using structured instruction design. Our approach supports broader and more fine-grained coverage of user preferences than existing methods. For preference-guided generation, we integrate global keyword-based control and local region-aware cross-attention modulation to steer the diffusion model without additional training, enabling precise alignment across both global attributes and local elements. The entire framework supports multi-round interactive refinement, facilitating real-time and context-aware image generation. Extensive experiments on the Viper dataset and our collected benchmark demonstrate that our method outperforms prior approaches in both quantitative metrics and human evaluations, and opens up new possibilities for dialog-based generation and MLLM-diffusion integration.

discussion (0)

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

Forward citations

Cited by 2 Pith papers

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

  1. Pseudo-Unification: Entropy Probing Reveals Divergent Information Patterns in Unified Multimodal Models

    cs.CV 2026-04 unverdicted novelty 7.0

    Unified multimodal models exhibit pseudo-unification due to modality-asymmetric entropy encoding and pattern-split responses between text and image generation.

  2. Premier: Personalized Preference Modulation with Learnable User Embedding in Text-to-Image Generation

    cs.CV 2026-03 unverdicted novelty 6.0

    Premier learns user-specific embeddings to modulate text-to-image generation, outperforming prior methods on preference alignment, text consistency, and expert ratings even with limited history.

Reference graph

Works this paper leans on

53 extracted references · 23 linked inside Pith · cited by 2 Pith papers

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    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 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Anthropic. 2024. The Claude 3 Model Family: Opus, Sonnet, Haiku

    2024

  4. [4]

    Avrahami, O.; Hayes, T.; Gafni, O.; Gupta, S.; Taigman, Y.; Parikh, D.; Lischinski, D.; Fried, O.; and Yin, X. 2023. Spatext: Spatio-textual representation for controllable image generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 18370--18380

    2023

  5. [5]

    Bai, S.; Chen, K.; Liu, X.; Wang, J.; Ge, W.; Song, S.; Dang, K.; Wang, P.; Wang, S.; Tang, J.; et al. 2025. Qwen2. 5-VL Technical Report. arXiv preprint arXiv:2502.13923

    Pith/arXiv arXiv 2025

  6. [6]

    Black, K.; Janner, M.; Du, Y.; Kostrikov, I.; and Levine, S. 2023. Training diffusion models with reinforcement learning. arXiv preprint arXiv:2305.13301

    Pith/arXiv arXiv 2023

  7. [7]

    Chatterjee, A.; Stan, G. B. M.; Aflalo, E.; Paul, S.; Ghosh, D.; Gokhale, T.; Schmidt, L.; Hajishirzi, H.; Lal, V.; Baral, C.; et al. 2025. Getting it right: Improving spatial consistency in text-to-image models. In European Conference on Computer Vision, 204--222. Springer

    2025

  8. [8]

    Chefer, H.; Alaluf, Y.; Vinker, Y.; Wolf, L.; and Cohen-Or, D. 2023. Attend-and-excite: Attention-based semantic guidance for text-to-image diffusion models. ACM Transactions on Graphics (TOG), 42(4): 1--10

    2023

  9. [9]

    Chen, M.; Laina, I.; and Vedaldi, A. 2024. Training-free layout control with cross-attention guidance. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 5343--5353

    2024

  10. [10]

    Chen, Z.; Zhang, L.; Weng, F.; Pan, L.; and Lan, Z. 2024. Tailored visions: Enhancing text-to-image generation with personalized prompt rewriting. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 7727--7736

    2024

  11. [11]

    Clark, K.; Vicol, P.; Swersky, K.; and Fleet, D. J. 2023. Directly Fine-Tuning Diffusion Models on Differentiable Rewards. arXiv:2309.17400

    Pith/arXiv arXiv 2023

  12. [12]

    Esser, P.; Kulal, S.; Blattmann, A.; Entezari, R.; M \"u ller, J.; Saini, H.; Levi, Y.; Lorenz, D.; Sauer, A.; Boesel, F.; et al. 2024. Scaling rectified flow transformers for high-resolution image synthesis. In Forty-first International Conference on Machine Learning

    2024

  13. [13]

    Fan, Y.; Watkins, O.; Du, Y.; Liu, H.; Ryu, M.; Boutilier, C.; Abbeel, P.; Ghavamzadeh, M.; Lee, K.; and Lee, K. 2024. Reinforcement learning for fine-tuning text-to-image diffusion models. Advances in Neural Information Processing Systems, 36

    2024

  14. [14]

    Frenkel, Y.; Vinker, Y.; Shamir, A.; and Cohen-Or, D. 2025. Implicit style-content separation using b-lora. In European Conference on Computer Vision, 181--198. Springer

    2025

  15. [15]

    Y.; Yang, Y.; and Gan, Z

    Fu, T.-J.; Hu, W.; Du, X.; Wang, W. Y.; Yang, Y.; and Gan, Z. 2023. Guiding instruction-based image editing via multimodal large language models. arXiv preprint arXiv:2309.17102

    Pith/arXiv arXiv 2023

  16. [16]

    H.; Chechik, G.; and Cohen-Or, D

    Gal, R.; Alaluf, Y.; Atzmon, Y.; Patashnik, O.; Bermano, A. H.; Chechik, G.; and Cohen-Or, D. 2022. An image is worth one word: Personalizing text-to-image generation using textual inversion. arXiv preprint arXiv:2208.01618

    Pith/arXiv arXiv 2022

  17. [17]

    Hertz, A.; Voynov, A.; Fruchter, S.; and Cohen-Or, D. 2024. Style aligned image generation via shared attention. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 4775--4785

    2024

  18. [18]

    Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33: 6840--6851

    2020

  19. [19]

    P.; Perelman, A.; Ramesh, A.; Clark, A.; Ostrow, A.; Welihinda, A.; Hayes, A.; Radford, A.; et al

    Hurst, A.; Lerer, A.; Goucher, A. P.; Perelman, A.; Ramesh, A.; Clark, A.; Ostrow, A.; Welihinda, A.; Hayes, A.; Radford, A.; et al. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276

    Pith/arXiv arXiv 2024

  20. [20]

    Kirstain, Y.; Polyak, A.; Singer, U.; Matiana, S.; Penna, J.; and Levy, O. 2023. Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation. arXiv:2305.01569

    Pith/arXiv arXiv 2023

  21. [21]

    Labs, B. F. 2024. FLUX: Inference Repository. https://github.com/black-forest-labs/flux. Accessed: 2024-10-25

    2024

  22. [22]

    Lee, S.; Kim, S.; Park, S.; Kim, G.; and Seo, M. 2024. Prometheus-vision: Vision-language model as a judge for fine-grained evaluation. In Findings of the Association for Computational Linguistics ACL 2024, 11286--11315

    2024

  23. [23]

    Li, Y.; Liu, H.; Wu, Q.; Mu, F.; Yang, J.; Gao, J.; Li, C.; and Lee, Y. J. 2023. Gligen: Open-set grounded text-to-image generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 22511--22521

    2023

  24. [24]

    Li, Y.; Yang, S.; Wang, W.; and Dong, J. 2024. Beyond inserting: Learning identity embedding for semantic-fidelity personalized diffusion generation. arXiv preprint arXiv:2402.00631

    Pith/arXiv arXiv 2024

  25. [25]

    Lian, L.; Li, B.; Yala, A.; and Darrell, T. 2023. Llm-grounded diffusion: Enhancing prompt understanding of text-to-image diffusion models with large language models. arXiv preprint arXiv:2305.13655

    Pith/arXiv arXiv 2023

  26. [26]

    Liu, N.; Li, S.; Du, Y.; Torralba, A.; and Tenenbaum, J. B. 2022. Compositional visual generation with composable diffusion models. In European Conference on Computer Vision, 423--439. Springer

    2022

  27. [27]

    Lu, C.; Zhou, Y.; Bao, F.; Chen, J.; Li, C.; and Zhu, J. 2022. Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps. Advances in Neural Information Processing Systems, 35: 5775--5787

    2022

  28. [28]

    Mou, C.; Wang, X.; Xie, L.; Wu, Y.; Zhang, J.; Qi, Z.; and Shan, Y. 2024. T2i-adapter: Learning adapters to dig out more controllable ability for text-to-image diffusion models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 4296--4304

    2024

  29. [29]

    L.; Pal, C

    Pernias, P.; Rampas, D.; Richter, M. L.; Pal, C. J.; and Aubreville, M. 2023. W \"u rstchen: An efficient architecture for large-scale text-to-image diffusion models. arXiv preprint arXiv:2306.00637

    Pith/arXiv arXiv 2023

  30. [30]

    Podell, D.; English, Z.; Lacey, K.; Blattmann, A.; Dockhorn, T.; M \"u ller, J.; Penna, J.; and Rombach, R. 2023. Sdxl: Improving latent diffusion models for high-resolution image synthesis. arXiv preprint arXiv:2307.01952

    Pith/arXiv arXiv 2023

  31. [31]

    Prabhudesai, M.; Goyal, A.; Pathak, D.; and Fragkiadaki, K. 2023. Aligning Text-to-Image Diffusion Models with Reward Backpropagation. arXiv:2310.03739

    Pith/arXiv arXiv 2023

  32. [32]

    W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al

    Radford, A.; Kim, J. W.; Hallacy, C.; Ramesh, A.; Goh, G.; Agarwal, S.; Sastry, G.; Askell, A.; Mishkin, P.; Clark, J.; et al. 2021. Learning transferable visual models from natural language supervision. In International conference on machine learning, 8748--8763. PMLR

    2021

  33. [33]

    Razavi, A.; Van den Oord, A.; and Vinyals, O. 2019. Generating diverse high-fidelity images with vq-vae-2. Advances in neural information processing systems, 32

    2019

  34. [34]

    Rombach, R.; Blattmann, A.; Lorenz, D.; Esser, P.; and Ommer, B. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 10684--10695

    2022

  35. [35]

    Ruiz, N.; Li, Y.; Jampani, V.; Pritch, Y.; Rubinstein, M.; and Aberman, K. 2023. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 22500--22510

    2023

  36. [36]

    Salehi, S.; Shafiei, M.; Yeo, T.; Bachmann, R.; and Zamir, A. 2024. ViPer: Visual Personalization of Generative Models via Individual Preference Learning. arXiv preprint arXiv:2407.17365

    Pith/arXiv arXiv 2024

  37. [37]

    Shen, X.; Zhang, R.; Zhao, X.; Zhu, J.; and Xiao, X. 2024. Pmg: Personalized multimodal generation with large language models. In Proceedings of the ACM Web Conference 2024, 3833--3843

    2024

  38. [38]

    Tang, Z.; Rybin, D.; and Chang, T.-H. 2023. Zeroth-order optimization meets human feedback: Provable learning via ranking oracles. arXiv preprint arXiv:2303.03751

    Pith/arXiv arXiv 2023

  39. [39]

    M.; Hauth, A.; Millican, K.; et al

    Team, G.; Anil, R.; Borgeaud, S.; Alayrac, J.-B.; Yu, J.; Soricut, R.; Schalkwyk, J.; Dai, A. M.; Hauth, A.; Millican, K.; et al. 2023. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805

    Pith/arXiv arXiv 2023

  40. [40]

    Von R \"u tte, D.; Fedele, E.; Thomm, J.; and Wolf, L. 2023. Fabric: Personalizing diffusion models with iterative feedback. arXiv preprint arXiv:2307.10159

    Pith/arXiv arXiv 2023

  41. [41]

    Wallace, B.; Dang, M.; Rafailov, R.; Zhou, L.; Lou, A.; Purushwalkam, S.; Ermon, S.; Xiong, C.; Joty, S.; and Naik, N. 2024. Diffusion model alignment using direct preference optimization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 8228--8238

    2024

  42. [42]

    Wang, H.; Spinelli, M.; Wang, Q.; Bai, X.; Qin, Z.; and Chen, A. 2024. Instantstyle: Free lunch towards style-preserving in text-to-image generation. arXiv preprint arXiv:2404.02733

    Pith/arXiv arXiv 2024

  43. [43]

    Wu, X.; Sun, K.; Zhu, F.; Zhao, R.; and Li, H. 2023. Human Preference Score: Better Aligning Text-to-Image Models with Human Preference. arXiv:2303.14420

    Pith/arXiv arXiv 2023

  44. [44]

    Wu, Y.; Zhou, X.; Ma, B.; Su, X.; Ma, K.; and Wang, X. 2024. Ifadapter: Instance feature control for grounded text-to-image generation. arXiv preprint arXiv:2409.08240

    Pith/arXiv arXiv 2024

  45. [45]

    Xu, J.; Liu, X.; Wu, Y.; Tong, Y.; Li, Q.; Ding, M.; Tang, J.; and Dong, Y. 2023. ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation. arXiv:2304.05977

    Pith/arXiv arXiv 2023

  46. [46]

    Yang, J.; Feng, J.; and Huang, H. 2024. EmoGen: Emotional Image Content Generation with Text-to-Image Diffusion Models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 6358--6368

    2024

  47. [47]

    Yang, J.; Huang, Q.; Ding, T.; Lischinski, D.; Cohen-Or, D.; and Huang, H. 2023. Emoset: A large-scale visual emotion dataset with rich attributes. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 20383--20394

    2023

  48. [48]

    Yang, L.; Yu, Z.; Meng, C.; Xu, M.; Ermon, S.; and Bin, C. 2024. Mastering text-to-image diffusion: Recaptioning, planning, and generating with multimodal llms. In Forty-first International Conference on Machine Learning

    2024

  49. [49]

    Ye, H.; Zhang, J.; Liu, S.; Han, X.; and Yang, W. 2023. Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models. arXiv preprint arXiv:2308.06721

    Pith/arXiv arXiv 2023

  50. [50]

    Yu, C.; Lu, H.; Gao, J.; Tan, Q.; Yang, X.; Wang, Y.; Wu, Y.; and Vinitsky, E. 2024. Few-shot in-context preference learning using large language models. arXiv preprint arXiv:2410.17233

    Pith/arXiv arXiv 2024

  51. [51]

    Zhang, L.; Rao, A.; and Agrawala, M. 2023. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 3836--3847

    2023

  52. [52]

    Zhang, S.; Wang, B.; Wu, J.; Li, Y.; Gao, T.; Zhang, D.; and Wang, Z. 2024. Learning Multi-dimensional Human Preference for Text-to-Image Generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 8018--8027

    2024

  53. [53]

    Zhang, Z.; Zhang, A.; Li, M.; Zhao, H.; Karypis, G.; and Smola, A. 2023. Multimodal chain-of-thought reasoning in language models. arXiv preprint arXiv:2302.00923

    Pith/arXiv arXiv 2023

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.