pith. sign in

arxiv: 2606.01858 · v1 · pith:YTUYKCPNnew · submitted 2026-06-01 · 💻 cs.CV

Polaris: Scaling Up Instruction-Guided Image Generation Towards Millions of Personalized Style Needs

Zhi-Kai Chen , Jun-Peng Jiang , Jun-Jie Tao , De-Chuan Zhan , Han-Jia Ye This is my paper

Pith reviewed 2026-06-28 15:04 UTC · model grok-4.3

classification 💻 cs.CV
keywords instruction-guided image generationmodel retrievalpersonalized stylesadapterscheckpointsscalable generationfine-tuned moduleswithout additional training
0
0 comments X

The pith

Polaris retrieves from 6500 checkpoints and 75000 adapters to generate personalized images matching user instructions without new training.

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

The paper introduces Polaris as a retrieval framework to meet growing demands for highly personalized image styles by drawing on an existing library of fine-tuned modules instead of training new models for each request. It indexes a large collection of checkpoints and adapters and automatically selects and combines the most relevant ones based on a user's input and instruction. The approach targets scalable, controllable outputs that align with instructions while avoiding the cost of repeated fine-tuning. This enables the system to handle millions of distinct style needs by exploiting the expanding ecosystem of pre-trained components.

Core claim

Polaris is an intelligent retrieval framework that automatically selects and integrates suitable models from the model library based on a user's instructions. The key insight is that harnessing such a massive and heterogeneous pool requires not only finding the most relevant modules among thousands of candidates, but also aligning them effectively for instruction-driven generation and editing. Polaris addresses this challenge by indexing over 6,500 checkpoints and 75,000 adapters, and retrieving the most relevant components given a user's input and instruction. In doing so, it delivers scalable, controllable, and well-aligned generation without any additional training.

What carries the argument

The retrieval framework that indexes and selects relevant pre-trained checkpoints and adapters from a heterogeneous library to match user instructions for image generation and editing.

If this is right

  • Users obtain personalized image generation for diverse requirements without incurring the cost of fine-tuning new models for each case.
  • The growing library of fine-tuned modules and adapters can be systematically exploited rather than training fresh models repeatedly.
  • Generation remains scalable and controllable while staying well-aligned with the given instructions.
  • New personalized style needs can be addressed by retrieval from the existing pool instead of additional training.

Where Pith is reading between the lines

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

  • Larger libraries could eventually cover an even wider range of user instructions as more modules are added over time.
  • The same retrieval idea might reduce the need for task-specific training in other generative domains.
  • Effective combination of retrieved modules could produce results that exceed the capabilities of any single component in the library.

Load-bearing premise

A retrieval mechanism can reliably identify and combine heterogeneous pre-trained modules to satisfy arbitrary user instructions without any fine-tuning or additional training steps.

What would settle it

A user instruction for a specific novel style where the retrieved modules produce outputs that fail to match the requested characteristics, even when the library contains components trained on related styles.

Figures

Figures reproduced from arXiv: 2606.01858 by De-Chuan Zhan, Han-Jia Ye, Jun-Jie Tao, Jun-Peng Jiang, Zhi-Kai Chen.

Figure 1
Figure 1. Figure 1: Although large models fine-tuned on trillions of tokens better understand diverse user instructions, such as style generation, they still exhibit significant blind spots. We introduce Polaris, a retrieval framework over the Stable Diffusion model library that automatically identifies and invokes the most relevant models, addressing these blind spots and enabling diverse, personalized image generation. Abst… view at source ↗
Figure 2
Figure 2. Figure 2: An overview of Polaris for instruction-guided model selection. The method is divided into two parts: (1) an Instruction Parser and Region Masker for adapting to image editing, and (2) a Multimodal Retriever for finding suitable checkpoints and LoRA adapters based on user requirements. The retrieved adapters and generated masks are combined to guide diffusion-based image generation. of instruction. Formally… view at source ↗
Figure 3
Figure 3. Figure 3: An overview of the tree rerank method. By modifying the attention mask of the LLM, we enable the model to evaluate multiple adapters simultaneously while generating only a single token per evaluation. a base checkpoint or a LoRA adapter—is embedded into a shared representation space that combines textual and visual information using a CLIP encoder. For each model x (checkpoint c or adapter l), we first con… view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of training and inference efficiency across methods. Our approach, Polaris, requires no additional training and achieves inference time between the two baselines. Furthermore, by incorporating our proposed LLM Tree Rerank strategy, Polaris attains a 1.50× speedup in reranking, highlighting its practical efficiency advantage. corresponding output (the generated image). Leveraging this process, we… view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative results on different tasks of GEDIT-BENCH (Liu et al., 2025). Polaris achieves notably higher instruction-following success rates and more faithful outputs than InstructP2P, effectively handling style transformations that remain challenging even after fine-tuning. Moreover, the knowledge coverage of Polaris is comparable to Bagel, a trillion-token scale pretrained model. captures nuanced user-s… view at source ↗
Figure 6
Figure 6. Figure 6: provides additional qualitative results on the User-Bench benchmark. For the style extraction task, our method produces outputs with clearer adherence to the input instructions and more consistent stylistic fidelity compared to both baselines. For the local style change task, our approach delivers more targeted edits while preserving the surrounding content, outperforming InstructP2P and achieving results … view at source ↗
read the original abstract

Users increasingly expect image generation models to quickly adapt to highly diverse and personalized requirements, such as producing images with distinctive styles or characteristics. Traditional approaches rely on fine-tuning, which is costly and difficult to scale. To cope with these limitations, the community has accumulated a growing library of fine-tuned modules and adapters, where each component targets specific generation needs and collectively serves as a foundation for handling new demands. This naturally raises a question: instead of repeatedly training new models, can we systematically exploit this expanding ecosystem to better fulfill user instructions? To this end, we present Polaris, an intelligent retrieval framework that automatically selects and integrates suitable models from the model library based on a user's instructions. The key insight is that harnessing such a massive and heterogeneous pool requires not only finding the most relevant modules among thousands of candidates, but also aligning them effectively for instruction-driven generation and editing. Polaris addresses this challenge by indexing over 6,500 checkpoints and 75,000 adapters, and retrieving the most relevant components given a user's input and instruction. In doing so, it delivers scalable, controllable, and well-aligned generation -- without any additional training.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper presents Polaris, a retrieval framework that indexes over 6,500 checkpoints and 75,000 adapters and automatically selects and integrates the most relevant components to fulfill arbitrary user instructions for image generation and editing, claiming to deliver scalable, controllable, and well-aligned outputs without any additional training or fine-tuning.

Significance. If the retrieval-plus-integration mechanism can be shown to work reliably, the approach would allow the community to exploit the growing ecosystem of fine-tuned models rather than retraining for each new personalized need, potentially reducing compute costs for instruction-guided generation at scale.

major comments (2)
  1. [Abstract] Abstract: the central claim that retrieval 'delivers ... well-aligned generation -- without any additional training' rests on the untested premise that an automatic selector can both identify the correct subset from a heterogeneous pool and combine the modules (via unspecified operator) without destructive interference; no retrieval representation, similarity metric, or merging procedure is described.
  2. [Abstract] Abstract: no evaluation protocol, dataset, or quantitative metric (e.g., alignment scores, user-study results, or comparison against fine-tuning baselines) is supplied to substantiate the 'well-aligned' and 'controllable' assertions, leaving the soundness of the core contribution unassessable from the provided text.
minor comments (1)
  1. [Title] Title claims 'Millions of Personalized Style Needs' while the abstract reports only 6,500 checkpoints + 75,000 adapters; clarify whether the indexed library is intended to grow to millions or whether the title is aspirational.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed comments on the abstract. We address each point below and indicate planned revisions to strengthen the presentation of the core claims.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that retrieval 'delivers ... well-aligned generation -- without any additional training' rests on the untested premise that an automatic selector can both identify the correct subset from a heterogeneous pool and combine the modules (via unspecified operator) without destructive interference; no retrieval representation, similarity metric, or merging procedure is described.

    Authors: The abstract is intentionally high-level. The full manuscript details the retrieval representation (Section 3.2), similarity metric based on instruction embedding alignment (Section 3.3), and the merging operator that combines selected checkpoints and adapters while mitigating interference via weighted fusion (Section 4.2). We will revise the abstract to briefly reference these components so the central claim is better grounded in the described method. revision: partial

  2. Referee: [Abstract] Abstract: no evaluation protocol, dataset, or quantitative metric (e.g., alignment scores, user-study results, or comparison against fine-tuning baselines) is supplied to substantiate the 'well-aligned' and 'controllable' assertions, leaving the soundness of the core contribution unassessable from the provided text.

    Authors: Section 5 of the manuscript presents the evaluation protocol, including the datasets used, quantitative metrics such as alignment scores and controllability measures, user studies, and direct comparisons against fine-tuning baselines. We will update the abstract to explicitly note that these results support the claims of well-aligned and controllable generation. revision: partial

Circularity Check

0 steps flagged

No circularity: framework proposal with no equations or self-referential derivations

full rationale

The provided abstract and context describe a retrieval-based system for selecting and integrating pre-trained models/adapters. No equations, fitted parameters, or derivation chains are present. Claims rest on the engineering premise of retrieval plus integration rather than any mathematical reduction to inputs. No self-citations, ansatzes, or uniqueness theorems are invoked in the given text. This matches the expected non-finding for systems papers without load-bearing formal steps.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no free parameters, axioms, or invented entities are specified in the provided text.

pith-pipeline@v0.9.1-grok · 5743 in / 1069 out tokens · 18214 ms · 2026-06-28T15:04:42.193643+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

95 extracted references · 3 canonical work pages

  1. [1]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Implicit diffusion models for continuous super-resolution , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  2. [2]

    , author=

    Lora: Low-rank adaptation of large language models. , author=. ICLR , volume=

  3. [3]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Adding conditional control to text-to-image diffusion models , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  4. [4]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    High-resolution image synthesis with latent diffusion models , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  5. [7]

    ACM computing surveys , volume=

    Diffusion models: A comprehensive survey of methods and applications , author=. ACM computing surveys , volume=. 2023 , publisher=

    2023

  6. [8]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Instructpix2pix: Learning to follow image editing instructions , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  7. [9]

    ACM SIGGRAPH 2022 conference proceedings , pages=

    Palette: Image-to-image diffusion models , author=. ACM SIGGRAPH 2022 conference proceedings , pages=

    2022

  8. [10]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Deadiff: An efficient stylization diffusion model with disentangled representations , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  9. [11]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Diffuse attend and segment: Unsupervised zero-shot segmentation using stable diffusion , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  10. [12]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  11. [13]

    Advances in Neural Information Processing Systems , volume=

    Stylus: Automatic adapter selection for diffusion models , author=. Advances in Neural Information Processing Systems , volume=

  12. [15]

    and Lo, Wan-Yen and Dollar, Piotr and Girshick, Ross , title =

    Kirillov, Alexander and Mintun, Eric and Ravi, Nikhila and Mao, Hanzi and Rolland, Chloe and Gustafson, Laura and Xiao, Tete and Whitehead, Spencer and Berg, Alexander C. and Lo, Wan-Yen and Dollar, Piotr and Girshick, Ross , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2023 , pages =

    2023

  13. [16]

    2025 , eprint=

    Qwen2.5-VL Technical Report , author=. 2025 , eprint=

    2025

  14. [17]

    Forty-first international conference on machine learning , year=

    Scaling rectified flow transformers for high-resolution image synthesis , author=. Forty-first international conference on machine learning , year=

  15. [18]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Reproducible scaling laws for contrastive language-image learning , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  16. [19]

    International conference on machine learning , pages=

    Zero-shot text-to-image generation , author=. International conference on machine learning , pages=. 2021 , organization=

    2021

  17. [21]

    arXiv preprint arXiv:2408.02657 , year =

    Lumina-mgpt: Illuminate flexible photorealistic text-to-image generation with multimodal generative pretraining , author =. arXiv preprint arXiv:2408.02657 , year =

    arXiv

  18. [22]

    2024 , eprint =

    Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation , author =. 2024 , eprint =

    2024

  19. [23]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Analyzing and improving the image quality of stylegan , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  20. [24]

    Proceedings of the IEEE international conference on computer vision , pages=

    Stackgan: Text to photo-realistic image synthesis with stacked generative adversarial networks , author=. Proceedings of the IEEE international conference on computer vision , pages=

  21. [27]

    arXiv preprint arXiv:2209.14491 , year=

    Re-imagen: Retrieval-augmented text-to-image generator , author=. arXiv preprint arXiv:2209.14491 , year=

    arXiv

  22. [28]

    2024 , eprint=

    Retrieval-Augmented Mixture of LoRA Experts for Uploadable Machine Learning , author=. 2024 , eprint=

    2024

  23. [31]

    Advances in Neural Information Processing Systems , volume=

    Model zoos: A dataset of diverse populations of neural network models , author=. Advances in Neural Information Processing Systems , volume=

  24. [32]

    2025 , eprint=

    Qwen2.5 Technical Report , author=. 2025 , eprint=

    2025

  25. [33]

    2024 , eprint=

    Text-to-image Diffusion Models in Generative AI: A Survey , author=. 2024 , eprint=

    2024

  26. [34]

    Extended abstracts of the 2021 CHI conference on human factors in computing systems , pages=

    Prompt programming for large language models: Beyond the few-shot paradigm , author=. Extended abstracts of the 2021 CHI conference on human factors in computing systems , pages=

    2021

  27. [35]

    2025 , eprint=

    Unified Multimodal Understanding and Generation Models: Advances, Challenges, and Opportunities , author=. 2025 , eprint=

    2025

  28. [40]

    arXiv preprint arXiv:2505.14683 , year=

    Emerging properties in unified multimodal pretraining , author=. arXiv preprint arXiv:2505.14683 , year=

    Pith/arXiv arXiv

  29. [42]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Open-vocabulary semantic segmentation with mask-adapted clip , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  30. [43]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Zone: Zero-shot instruction-guided local editing , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  31. [45]

    2015 , journal =

    An Empirical Investigation of Catastrophic Forgetting in Gradient-Based Neural Networks , author=. 2015 , journal =

    2015

  32. [46]

    2025 , journal =

    Overcoming catastrophic forgetting in neural networks , author=. 2025 , journal =

    2025

  33. [47]

    2022 , journal =

    Gradient Episodic Memory for Continual Learning , author=. 2022 , journal =

    2022

  34. [48]

    Journal of Big Data , year=

    Survey on deep learning with class imbalance , author=. Journal of Big Data , year=

  35. [49]

    2022 , journal =

    Self-Supervised Aggregation of Diverse Experts for Test-Agnostic Long-Tailed Recognition , author=. 2022 , journal =

    2022

  36. [51]

    2023 , journal =

    A Survey of Diffusion Based Image Generation Models: Issues and Their Solutions , author=. 2023 , journal =

    2023

  37. [52]

    2024 , journal =

    Rich Human Feedback for Text-to-Image Generation , author=. 2024 , journal =

    2024

  38. [53]

    2025 , journal =

    DreamOmni: Unified Image Generation and Editing , author=. 2025 , journal =

    2025

  39. [54]

    2025 , journal =

    Diffusion-4K: Ultra-High-Resolution Image Synthesis with Latent Diffusion Models , author=. 2025 , journal =

    2025

  40. [55]

    2025 , journal =

    Personalized Image Generation with Deep Generative Models: A Decade Survey , author=. 2025 , journal =

    2025

  41. [56]

    A Survey of AI Text-to-Image and AI Text-to-Video Generators , DOI=

    Singh, Aditi , year=. A Survey of AI Text-to-Image and AI Text-to-Video Generators , DOI=. 2023 4th International Conference on Artificial Intelligence, Robotics and Control (AIRC) , publisher=

    2023

  42. [57]

    2023 , journal =

    BLIP-Diffusion: Pre-trained Subject Representation for Controllable Text-to-Image Generation and Editing , author=. 2023 , journal =

    2023

  43. [59]

    2024 , journal =

    MM-LLMs: Recent Advances in MultiModal Large Language Models , author=. 2024 , journal =

    2024

  44. [60]

    2020 , journal =

    Denoising Diffusion Probabilistic Models , author=. 2020 , journal =

    2020

  45. [61]

    2021 , journal =

    Score-Based Generative Modeling through Stochastic Differential Equations , author=. 2021 , journal =

    2021

  46. [62]

    2022 , journal =

    Towards a Unified View of Parameter-Efficient Transfer Learning , author=. 2022 , journal =

    2022

  47. [63]

    2024 , journal =

    Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey , author=. 2024 , journal =

    2024

  48. [64]

    2024 , journal =

    GPT-4o System Card , author=. 2024 , journal =

    2024

  49. [65]

    Qwen2.5-vl technical report, 2025

    Bai, S., Chen, K., Liu, X., Wang, J., Ge, W., Song, S., Dang, K., Wang, P., Wang, S., Tang, J., Zhong, H., Zhu, Y., Yang, M., Li, Z., Wan, J., Wang, P., Ding, W., Fu, Z., Xu, Y., Ye, J., Zhang, X., Xie, T., Cheng, Z., Zhang, H., Yang, Z., Xu, H., and Lin, J. Qwen2.5-vl technical report, 2025

    2025

  50. [66]

    Brooks, T., Holynski, A., and Efros, A. A. Instructpix2pix: Learning to follow image editing instructions. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 18392--18402, 2023 a

    2023

  51. [67]

    Brooks, T., Holynski, A., and Efros, A. A. Instructpix2pix: Learning to follow image editing instructions. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 18392--18402, 2023 b

    2023

  52. [68]

    Scaling rectified flow transformers for high-resolution image synthesis

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

    2024

  53. [69]

    A model zoo of vision transformers

    Falk, D., Meynent, L., Pfammatter, F., Sch \"u rholt, K., and Borth, D. A model zoo of vision transformers. arXiv preprint arXiv:2504.10231, 2025

    arXiv 2025

  54. [70]

    Implicit diffusion models for continuous super-resolution

    Gao, S., Liu, X., Zeng, B., Xu, S., Li, Y., Luo, X., Liu, J., Zhen, X., and Zhang, B. Implicit diffusion models for continuous super-resolution. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 10021--10030, 2023

    2023

  55. [71]

    J., Mirza, M., Xiao, D., Courville, A., and Bengio, Y

    Goodfellow, I. J., Mirza, M., Xiao, D., Courville, A., and Bengio, Y. An empirical investigation of catastrophic forgetting in gradient-based neural networks. arXiv preprint arXiv:1312.6211, 2015

    Pith/arXiv arXiv 2015

  56. [72]

    Ha, D., Dai, A., and Le, Q. V. Hypernetworks. arXiv preprint arXiv:1609.09106, 2016

    Pith/arXiv arXiv 2016

  57. [73]

    Han, Z., Gao, C., Liu, J., Zhang, J., and Zhang, S. Q. Parameter-efficient fine-tuning for large models: A comprehensive survey, 2024

    2024

  58. [74]

    Towards a unified view of parameter-efficient transfer learning

    He, J., Zhou, C., Ma, X., Berg-Kirkpatrick, T., and Neubig, G. Towards a unified view of parameter-efficient transfer learning. arXiv preprint arXiv:2110.04366, 2022

    arXiv 2022

  59. [75]

    L., and Choi, Y

    Hessel, J., Holtzman, A., Forbes, M., Bras, R. L., and Choi, Y. Clipscore: A reference-free evaluation metric for image captioning. arXiv preprint arXiv:2104.08718, 2021

    Pith/arXiv arXiv 2021

  60. [76]

    Denoising diffusion probabilistic models

    Ho, J., Jain, A., and Abbeel, P. Denoising diffusion probabilistic models. arXiv preprint arXiv:2006.11239, 2020

    Pith/arXiv arXiv 2006

  61. [77]

    J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., Chen, W., et al

    Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., Chen, W., et al. Lora: Low-rank adaptation of large language models. ICLR, 1 0 (2): 0 3, 2022

    2022

  62. [78]

    Y., Pang, T., Du, C., and Lin, M

    Huang, C., Liu, Q., Lin, B. Y., Pang, T., Du, C., and Lin, M. Lorahub: Efficient cross-task generalization via dynamic lora composition. arXiv preprint arXiv:2307.13269, 2023

    arXiv 2023

  63. [79]

    Unified language-vision pretraining in llm with dynamic discrete visual tokenization

    Jin, Y., Xu, K., Chen, L., Liao, C., Tan, J., Huang, Q., Chen, B., Lei, C., Liu, A., Song, C., et al. Unified language-vision pretraining in llm with dynamic discrete visual tokenization. arXiv preprint arXiv:2309.04669, 2023

    arXiv 2023

  64. [80]

    Johnson, J. M. and Khoshgoftaar, T. M. Survey on deep learning with class imbalance. Journal of Big Data, 2019

    2019

  65. [81]

    C., Lo, W.-Y., Dollar, P., and Girshick, R

    Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A. C., Lo, W.-Y., Dollar, P., and Girshick, R. Segment anything. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pp.\ 4015--4026, October 2023

    2023

  66. [82]

    Zone: Zero-shot instruction-guided local editing

    Li, S., Zeng, B., Feng, Y., Gao, S., Liu, X., Liu, J., Li, L., Tang, X., Hu, Y., Liu, J., et al. Zone: Zero-shot instruction-guided local editing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 6254--6263, 2024

    2024

  67. [83]

    Open-vocabulary semantic segmentation with mask-adapted clip

    Liang, F., Wu, B., Dai, X., Li, K., Zhao, Y., Zhang, H., Zhang, P., Vajda, P., and Marculescu, D. Open-vocabulary semantic segmentation with mask-adapted clip. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 7061--7070, 2023

    2023

  68. [84]

    Step1x-edit: A practical framework for general image editing

    Liu, S., Han, Y., Xing, P., Yin, F., Wang, R., Cheng, W., Liao, J., Wang, Y., Fu, H., Han, C., et al. Step1x-edit: A practical framework for general image editing. arXiv preprint arXiv:2504.17761, 2025

    Pith/arXiv arXiv 2025

  69. [85]

    Loke, B. S. Y., Quadri, F., Vivanco, G., Casagrande, M., and Fenollosa, S. Overcoming catastrophic forgetting in neural networks. arXiv preprint arXiv:2507.10485, 2025

    arXiv 2025

  70. [86]

    and Ranzato, M

    Lopez-Paz, D. and Ranzato, M. Gradient episodic memory for continual learning. arXiv preprint arXiv:1706.08840, 2022

    arXiv 2022

  71. [87]

    E., Salakhutdinov, R., Stoica, I., et al

    Luo, M., Wong, J., Trabucco, B., Huang, Y., Gonzalez, J. E., Salakhutdinov, R., Stoica, I., et al. Stylus: Automatic adapter selection for diffusion models. Advances in Neural Information Processing Systems, 37: 0 32888--32915, 2024

    2024

  72. [88]

    Michelessa, M., Ng, J., Hurter, C., and Lim, B. Y. Varif.ai to vary and verify user-driven diversity in scalable image generation. In Proceedings of the 2025 ACM Designing Interactive Systems Conference, DIS ’25, pp.\ 1867–1885. ACM, July 2025. doi:10.1145/3715336.3735847. URL http://dx.doi.org/10.1145/3715336.3735847

    work page doi:10.1145/3715336.3735847 2025

  73. [89]

    Gpt-4o system card

    OpenAI et al. Gpt-4o system card. arXiv preprint arXiv:2410.21276, 2024

    Pith/arXiv arXiv 2024

  74. [90]

    Sdxl: Improving latent diffusion models for high-resolution image synthesis

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

    Pith/arXiv arXiv 2023

  75. [91]

    Deadiff: An efficient stylization diffusion model with disentangled representations

    Qi, T., Fang, S., Wu, Y., Xie, H., Liu, J., Chen, L., He, Q., and Zhang, Y. Deadiff: An efficient stylization diffusion model with disentangled representations. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 8693--8702, 2024

    2024

  76. [92]

    and Chaudhari, P

    Ramesh, R. and Chaudhari, P. Model zoo: A growing" brain" that learns continually. arXiv preprint arXiv:2106.03027, 2021

    arXiv 2021

  77. [93]

    High-resolution image synthesis with latent diffusion models

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

    2022

  78. [94]

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Ruiz, N., Li, Y., Jampani, V., Pritch, Y., Rubinstein, M., and Aberman, K. 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, pp.\ 22500--22510, 2023 a

    2023

  79. [95]

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Ruiz, N., Li, Y., Jampani, V., Pritch, Y., Rubinstein, M., and Aberman, K. 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, pp.\ 22500--22510, 2023 b

    2023

  80. [96]

    Palette: Image-to-image diffusion models

    Saharia, C., Chan, W., Chang, H., Lee, C., Ho, J., Salimans, T., Fleet, D., and Norouzi, M. Palette: Image-to-image diffusion models. In ACM SIGGRAPH 2022 conference proceedings, pp.\ 1--10, 2022

    2022

Showing first 80 references.