Pith. sign in

REVIEW 4 major objections 5 minor 52 references

In-Context Brush: Zero-shot Customized Subject Insertion with Context-Aware Latent Space Manipulation

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read The paper proposes a training-free framework that inserts a user-specified subject into a masked image region by treating the reference photo plus text prompt as in-context demonstrations and shifting attention hidden states in a…

desk verdict A useful training-free subject insertion recipe with broad empirical support, but the theoretical derivation in Sec 3.2 is mis-specified and the lack of code/data leaves the reported gains unverifiable. read the letter →

arxiv 2505.20271 v1 pith:32O4P2IX submitted 2025-05-26 cs.CV cs.AIcs.GR

classification cs.CVcs.AIcs.GR
keywords customizedsubjectinsertionzero-shotimageeditingin-contextlearningdiffusiontransformerlatentfeatureshiftingattentionreweightinginpaintingtraining-freegeneration
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

In-Context Brush reformulates customized subject insertion as an in-context learning problem: the reference subject image and the text prompt act as demonstrations, and the target image with the masked region acts as the query. The paper argues that inside a pretrained multi-modal diffusion transformer, the hidden state of query tokens is already a weighted mixture of prompt, subject, and query attention contributions, so identity transfer can be achieved by amplifying those contributions with two scalar strengths, plus reweighting attention heads for prompt control and blending tokens to preserve background consistency. The claimed payoff is superior identity preservation, text alignment, and image quality compared with eight state-of-the-art baselines, with no training and no extra data collection. If true, this would make customized insertion a test-time operation on an off-the-shelf inpainting model.

What carries the argument

The load-bearing object is Eq. (5), the latent feature shift: $h_s \leftarrow h_s + \alpha_1 A_{s,p} v_p + \alpha_2 A_{s,c} v_c$, applied to the attention heads of an MMDiT-based inpainting model. It says that reference-subject and prompt information can be injected into the query hidden state by adding weighted attention-map-by-value products that are already present in the joint self-attention computation. Two auxiliary mechanisms carry the rest of the pipeline: head-wise reweighting, which normalizes prompt-to-query attention activation per head and scales the query hidden state to amplify prompt-relevant heads, and token blending, which replaces background tokens with the noised ground-truth background at each step to prevent distribution drift.

What would settle it

Run the method on a held-out set of subject/scene pairs whose content differs from the ablation set, fixing both shift strengths at the ablation-chosen values. If identity (DINO) or text-alignment (CLIP-T) scores fall well below the reported means, or visible artifacts appear on many pairs, the claim that a fixed latent shift generalizes across inputs is falsified. A sharper test is to search for the per-image best strengths: if the maximizing pair varies strongly across images, the training-free claim depends on per-image tuning rather than a fixed mechanism.

Watch

Extended reading notes

Core claim

The central claim is that subject-level in-context transfer can be realized as a shift of attention hidden states at test time. The authors derive that $h_s$, the hidden state belonging to the query region, decomposes into $\alpha_p h_{\text{demo}_p} + \alpha_c h_{\text{demo}_c} + \alpha_s h_{\text{query}}$, and then propose the feature shift injection $h_s \leftarrow h_s + \alpha_1 A_{s,p} v_p + \alpha_2 A_{s,c} v_c$ to push reference identity and prompt semantics into the output. With head-wise reweighting that scales query hidden states by normalized prompt-attention activation, and token blending that re-fuses the unmasked background at every denoising step, the method claims to match or beat training-based and training-free baselines on DINO, CLIP-I, CLIP-T, and FID without updating any weights.

Load-bearing premise

The method assumes that amplifying the prompt- and reference-attention contributions already present in the model's internal states, with two hand-set strengths, improves identity and prompt alignment without introducing artifacts.

Editorial extensions

If this is right

  • Per-subject fine-tuning and paired training data become unnecessary: a single reference photo and a mask are enough to insert that object into a new scene on an off-the-shelf MMDiT inpainting model.
  • Prompt-driven editing of the inserted subject is achievable at test time, because head-wise reweighting amplifies attention heads that respond to the prompt tokens; materials, colors, and poses can be changed without retraining.
  • The same latent-shift operation can be reused for virtual try-on, compositional scene building, and partial part insertion, since all three are mask-specified placements of reference content into a target.
  • Background consistency is protected by re-injecting the unmasked target each denoising step, which avoids the distribution drift that otherwise causes seams and tone mismatches.

Reading between the lines

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

  • Because the shift strengths $\alpha_1$ and $\alpha_2$ are hand-set with no stated default, an implicit next step is to predict them from the reference and target images; if the optimal pair varies across inputs, the method would need per-image tuning rather than a single fixed recipe.
  • The mechanism only amplifies contributions the base model already computes, so it should compose with other attention-level edits such as prompt-to-prompt style changes, as long as those edits also target attention maps.
  • Token blending is a general anti-drift regularizer for multi-step latent compositing and could improve other methods that concatenate reference and target latents, though the paper only demonstrates it inside this pipeline.
  • Strongly increasing $\alpha_2$ is likely to trade editability for identity fidelity: the better the reference appearance is preserved, the harder it becomes to change intrinsic attributes such as material or color.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes In-Context Brush, a training-free method for customized subject insertion that concatenates a reference subject image and a text prompt as in-context demonstrations with a target image as query, then manipulates the hidden states of a pretrained MMDiT-based inpainting model at test time. Three mechanisms are introduced: latent feature shifting (Sec. 3.2), head-wise attention reweighting (Sec. 3.3), and token blending (Sec. 3.4). The method is evaluated against eight baselines plus two-stage combinations on DINO, CLIP-I, CLIP-T, and FID, with a user study and ablations, and is claimed to achieve state-of-the-art identity preservation, prompt alignment, and image quality.

Significance. If the mechanism were correctly specified and reproducible, the work would be a useful contribution to training-free subject insertion, a practically relevant problem. The empirical evaluation is substantial: multiple baselines, a user study with 6,825 votes, ablation studies for each proposed module, and robustness checks over ten random seeds. The paper also honestly reports a limitation with similar contextual features in Sec. 6. However, the core derivation in Sec. 3.2 does not establish the proposed operation, the key strength parameters are not given default values, and no code is released, so the reported results cannot currently be traced to the described method. The central claim is plausible but not established as presented.

major comments (4)
  1. [Sec. 3.2, Eq. (5)] The operation in Eq. (5) is not entailed by the paper's own decomposition in Eq. (12). With the definitions of As,p, As,c, h(demo p), and h(demo c), one has As,p vp = αp·h(demo p) and As,c vc = αc·h(demo c). Substituting into Eq. (5) gives ĥs = hs + α1·αp·h(demo p) + α2·αc·h(demo c) = αs·h(query) + (1+α1)·αp·h(demo p) + (1+α2)·αc·h(demo c). Thus Eq. (5) double-counts demo terms already present in hs and changes their coefficients multiplicatively, not additively as claimed in the text. Because αp and αc are data-dependent row sums of the full softmax, they vary by token, denoising step, prompt length, and image token count, so the hand-set strengths α1 and α2 in Fig. 6 are not transferable in the way the paper states. The claimed 'proof' of latent feature shifting is therefore a rewriting of the attention equation plus an asserted operation, not a derivation of Eq. (5).
  2. [Sec. 3.2 and Sec. 4.5 (hyperparameters)] No default values for α1 and α2 are stated anywhere in the paper or supplementary, yet Table 1 reports quantitative results that depend on these values. The only numeric coefficient in the supplementary, 'The attention reweighting coefficient is set to 1.3' (Supp. C), is not tied to Eq. (5) and appears to belong to the Diptych baseline description. Moreover, the main text says that 'excessively large values of α1 and α2 (e.g., 0.5)' degrade quality, while Fig. 6 shows α=0.5 as the middle of the tested range; this is internally inconsistent and suggests the implemented operation may differ from the written Eq. (5). Without a specification of the implemented normalization and the α values used for Table 1, the reported gains cannot be reproduced.
  3. [Sec. 3.3, Eqs. (6)-(8)] The head-wise reweighting mechanism is under-specified and the motivation does not match the equation. The text states that 'we leverage h(demo p) to soft activate h(query)', but Eq. (6) computes the activation V_h from the attention block Ap,s, which is the prompt-to-subject attention map (xp query attending to xs keys). This is a different quantity from h(demo p) in Eq. (12), which is Attn(xsWq, xpWk, xpWv), i.e., subject attending to prompt. The paper provides no justification for why activation of Ap,s should be a reliable proxy for prompt-relevant heads, and Eq. (8) scales the entire per-head query hidden state without any renormalization, so the operation changes output magnitudes in an uncontrolled way. The proxy assumption and the notation need to be aligned and empirically validated.
  4. [Sec. 4.1, Dataset description] The dataset sizes are arithmetically inconsistent. The text says 30 subject images from DreamBooth plus 50 additional subject images from the Internet, and then states the evaluation dataset contains 100 subject images; 30+50=80, not 100. The scene counts (50 COCO + 80 Internet = 130) are consistent. This should be corrected because it affects the interpretation of the FID and other statistics reported on this dataset.
minor comments (5)
  1. [Eq. (2) and Supp. Eq. (10)] There is a typo: 'Concatente' should be 'Concatenate'.
  2. [Table 1] The formatting of the CLIP-T entries for 'Ours w/o head' is broken: '0 .2682±0.331' contains a stray space, and the standard deviation 0.331 is implausibly large compared with 0.0365 for the full method; this likely reflects a transcription error.
  3. [Sec. 4.5, Fig. 6] The statement that α1=0.5 is 'excessively large' is hard to reconcile with Fig. 6, where α1=0.5 is a displayed setting; please clarify which values are recommended and which were used in the main results.
  4. [Sec. 4.3, FID] The FID is computed on a small evaluation set (on the order of 100 images), which makes the FID differences in Table 1 difficult to interpret; reporting confidence intervals or additional image-quality metrics would strengthen the claim.
  5. [Supp. C] The sentence 'The attention reweighting coefficient is set to 1.3' is ambiguous: specify whether this coefficient applies to the Diptych baseline or to the proposed method, and if it is part of the proposed method, explain how it relates to α1 and α2.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the central claim is an external benchmark comparison; the only self-citation is non-load-bearing motivation, while the Eq. 5 derivation has an internal arithmetic inconsistency that is a correctness/reproducibility concern rather than a circularity.

full rationale

The paper's strongest claim is empirical: superior identity preservation, text alignment, and image quality versus eight baselines, evaluated with DINO, CLIP-I, CLIP-T, FID, and a user study in Tables 1-2 and Figures 3-5. These evaluations are independent of the method's derivation, so the main result is not circular. The core derivation in Sec. 3.2 rewrites the joint-attention output as hs = αp·h(demo p) + αc·h(demo c) + αs·h(query) and then proposes Eq. 5, hs_hat = hs + α1·As,p·vp + α2·As,c·vc, as a 'latent feature shift.' That is an intervention heuristic, not a prediction forced by the derivation. I checked the skeptic's concern: under the paper's own definitions, As,p·vp = αp·h(demo p) and As,c·vc = αc·h(demo c), because As,p is the prompt block of the full softmax row. Substituting into Eq. 5 gives hs_hat = αs·h(query) + (1+α1)·αp·h(demo p) + (1+α2)·αc·h(demo c). Thus the text's claim that Eq. 5 'directly amplifies the values of scalars αp and αc' is arithmetically inaccurate: the coefficients become (1+α1)·αp and (1+α2)·αc, not αp+α1 and αc+α2, and the terms already present in hs are double-counted. This is an internal formal inconsistency and an under-specification of the effective operation, not a circularity: the benchmark results do not rely on the correctness of that derivation. The paper also does not state fixed default values for α1 and α2, and the only concrete coefficient in the supplementary (1.3) is attached to the Diptych comparison; this plus the absence of released code is a reproducibility risk, not evidence that a fitted parameter was renamed as a prediction. The only self-citation is [47] (Head-Router, with overlapping authors), used in Sec. 3.3 to motivate the idea that attention heads specialize semantically. That citation is not load-bearing: the head-reweighting module's effect is independently ablated in Fig. 8 and Table 1, and the central identity-preservation claim does not reduce to it. Accordingly, I find no circular step that rises to the level of a fitted-input-called-prediction or self-citation-chain dependence; the appropriate score is 2 due to the minor, non-load-bearing self-citation, with the Eq. 5 issue flagged as a correctness concern rather than circularity.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The method contributes no new physical or conceptual entities; it relies on existing pretrained models and hand-set inference-time parameters. The most load-bearing postulates are that the base MMDiT can encode cross-modal subject semantics in attention, and that amplifying demo attention terms and reweighting heads improves fidelity and prompt control.

free parameters (2)
  • α1 (text shift strength) = No fixed default reported; ablation tests 0.1, 0.5, 1.5
    Controls how strongly prompt-derived attention outputs are added to the query hidden states in Eq 5; chosen by hand per use, not derived.
  • α2 (subject shift strength) = No fixed default reported; ablation tests 0.1, 0.5, 1.5
    Controls how strongly reference-subject attention outputs are added to the query hidden states in Eq 5; hand-tuned.
assumptions (6)
  • standard math Joint attention in MMDiT can be decomposed into prompt, reference, and query components via Eq 10/12.
    Used in Sec 3.2 to rewrite hs and justify feature shifting; correct by linear algebra of attention.
  • domain assumption Flux-1.0-fill[dev] pretrained inpainting model can transfer subject semantics across concatenated reference/query inputs when its attention maps are manipulated.
    The entire method assumes the base model has enough latent subject knowledge; no fine-tuning is performed to inject the subject.
  • ad hoc to paper Amplifying demo attention contributions with hand-set α1/α2 yields faithful insertion without artifacts.
    The core heuristic Eq 5; effectiveness is validated only empirically, not derived.
  • ad hoc to paper The activation of prompt-to-query attention map Ap,s per head is a reliable proxy for which heads to reward for prompt alignment.
    Used in Eq 6-8; no theoretical guarantee that maximizing this activation corresponds to prompt faithfulness.
  • domain assumption Grounding DINO and SAM segment the subject accurately in all benchmark images.
    The input reference subject is extracted automatically; errors would propagate into generation.
  • domain assumption Token blending (noising the unmasked background at each step) preserves unbiased background semantics.
    Eq 9 assumes that re-injecting the original background latent each step avoids distribution drift; standard inpainting practice but not proven here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of In-Context Brush: Zero-shot Customized Subject Insertion with Context-Aware Latent Space Manipulation." pith.science (2026). https://pith.science/paper/32O4P2IX

@misc{pith2026250520271,
  author       = {Pith},
  title        = {Pith review of: In-Context Brush: Zero-shot Customized Subject Insertion with Context-Aware Latent Space Manipulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/32O4P2IX}},
  note         = {Machine review of arXiv:2505.20271}
}
read the original abstract

Recent advances in diffusion models have enhanced multimodal-guided visual generation, enabling customized subject insertion that seamlessly "brushes" user-specified objects into a given image guided by textual prompts. However, existing methods often struggle to insert customized subjects with high fidelity and align results with the user's intent through textual prompts. In this work, we propose "In-Context Brush", a zero-shot framework for customized subject insertion by reformulating the task within the paradigm of in-context learning. Without loss of generality, we formulate the object image and the textual prompts as cross-modal demonstrations, and the target image with the masked region as the query. The goal is to inpaint the target image with the subject aligning textual prompts without model tuning. Building upon a pretrained MMDiT-based inpainting network, we perform test-time enhancement via dual-level latent space manipulation: intra-head "latent feature shifting" within each attention head that dynamically shifts attention outputs to reflect the desired subject semantics and inter-head "attention reweighting" across different heads that amplifies prompt controllability through differential attention prioritization. Extensive experiments and applications demonstrate that our approach achieves superior identity preservation, text alignment, and image quality compared to existing state-of-the-art methods, without requiring dedicated training or additional data collection.

Figures

Figures reproduced from arXiv: 2505.20271 by the authors.

Figure 1
Figure 1. Our method achieves identity-preserving subject insertion in the novel scene harmoniously, simultaneously enabling diverse [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Pipeline of our method. We mainly introduce latent space shifting for subject present in target images in a training-free manner. In the “Latent Feature Shifting” part, features from the reference are shifted to output. We propose attention heads activation for further enhance representation of textual prompts and token blending for consistency injection within the image. image structure, enabling the model to effec… view at source ↗
Figure 3
Figure 3. Qualitative comparison on subject injection and editing with baseline methods. Results of our results maintain identity consistency with reference while preserving fine-grained features, and are also aligning with the prompts. Masks are labeled as white boxes on target images. collect 50 diverse scenes as target images from COCO dataset [26]. We also collected 50 additional subject images and 80 scene images from th… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: User study results. evaluation index results and their standard deviations. As show in Tab. 1, our method outperforms baselines in subject identity alignment, editing alignment, and image quality. Although the editing score (third column) is slightly lower than Break-a…
Figure 5
Figure 5. Figure 5: Comparisons with two-stage methods. parameters based on the specific image to balance the control of the image and generation quality. 𝛼1=0.1 𝛼1=0.5 𝛼1=1.5 𝛼2=0.1 𝛼2=0.5 𝛼2=1.5 Reference “A teapot in metal material” [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Ablation study on shift strength. Customized subject insertion via basic Flux-Fill model We build an inpainting pipeline with Ic and Is concatenated as input to evaluate the basic customized subject insertion ability of Flux￾Fill. Results in [PITH_FULL_IMAGE:figures/f…
Figure 7
Figure 7. Figure 7: Ablation study on token blending. Without token blending We ablate the token blending module and present results in [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Ablation study on attention heads activation. of latents from the target and fusion with latents from subjects along the denoising step, the presentation of subjects in the target images has better interaction with the background, achieving the overall consistency of t…
Figure 10
Figure 10. Figure 10: Application of compositional generation. Car wheels Legs of the chair [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]
Figure 11
Figure 11. Figure 11: Application of partly insertion. Partly insertion. Our method supports selective part-based insertion, enabling users to transfer specific regions from reference into corresponding locations of the generated output. This allows fine-grained control while maintaining s…
Figure 12
Figure 12. Figure 12: Badcase. In some cases, interaction among similar contextual features in attention calculation may cause same features in appearance of results as the concepts from background. 7. Conclusions In this work, we leverage ICL to activate the context-consistent generation …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 37 canonical work pages

  1. [1]

    Gpt-4 technical report

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023. 2

  2. [2]

    Break-a-scene: Extracting multiple concepts from a single image

    Omri Avrahami, Kfir Aberman, Ohad Fried, Daniel Cohen- Or, and Dani Lischinski. Break-a-scene: Extracting multiple concepts from a single image. In SIGGRAPH Asia 2023 Conference Papers, pages 1–12, 2023. 3, 5

  3. [3]

    Flux, offering state-of-the-art performance image generation

    blackforestlabs.ai. Flux, offering state-of-the-art performance image generation. https://blackforestlabs.ai/,

  4. [4]

    Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901, 2020

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901, 2020. 2

  5. [5]

    Pixart- $\alpha$: Fast training of diffusion transformer for photorealistic text-to-image syn- thesis

    Junsong Chen, Jincheng YU, Chongjian GE, Lewei Yao, Enze Xie, Zhongdao Wang, James Kwok, Ping Luo, Huchuan Lu, and Zhenguo Li. Pixart- $\alpha$: Fast training of diffusion transformer for photorealistic text-to-image syn- thesis. In The Twelfth International Conference on Learning Representations, 2024. 2

  6. [6]

    Improving in-context learning in diffusion models with visual context-modulated prompts

    Tianqi Chen, Yongfei Liu, Zhendong Wang, Jianbo Yuan, Quanzeng You, Hongxia Yang, and Mingyuan Zhou. Improving in-context learning in diffusion models with visual context-modulated prompts. arXiv preprint arXiv:2312.01408, 2023. 2

  7. [7]

    Zero-shot image editing with reference imitation

    Xi Chen, Yutong Feng, Mengting Chen, Yiyang Wang, Shilong Zhang, Yu Liu, Yujun Shen, and Hengshuang Zhao. Zero-shot image editing with reference imitation. Advances in Neural Information Processing Systems, 37:84010–84032,

  8. [8]

    Anydoor: Zero-shot object-level image customization

    Xi Chen, Lianghua Huang, Yu Liu, Yujun Shen, Deli Zhao, and Hengshuang Zhao. Anydoor: Zero-shot object-level image customization. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6593–6602, 2024. 2, 3, 8

Show all 52 references
  1. [9]

    Custom-edit: Text-guided image editing with customized diffusion models

    Jooyoung Choi, Yunjey Choi, Yunji Kim, Junho Kim, and Sungroh Yoon. Custom-edit: Text-guided image editing with customized diffusion models. arXiv preprint arXiv:2305.15779, 2023. 2

  2. [10]

    Turboedit: Text-based image editing using few-step diffusion models

    Gilad Deutch, Rinon Gal, Daniel Garibi, Or Patashnik, and Daniel Cohen-Or. Turboedit: Text-based image editing using few-step diffusion models. In SIGGRAPH Asia 2024 Conference Papers, pages 1–12, 2024. 5, 8

  3. [11]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805, 2018. 2

  4. [12]

    A survey on in-context learning

    Qingxiu Dong, Lei Li, Damai Dai, Ce Zheng, Jingyuan Ma, Rui Li, Heming Xia, Jingjing Xu, Zhiyong Wu, Tianyu Liu, et al. A survey on in-context learning. arXiv preprint arXiv:2301.00234, 2022. 2

  5. [13]

    Scaling rectified flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas M ¨uller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling rectified flow transformers for high-resolution image synthesis. In Forty-first International Conference on Machin...

  6. [14]

    An image is worth one word: Personalizing text-to- image generation using textual inversion

    Rinon Gal, Yuval Alaluf, Yuval Atzmon, Or Patashnik, Amit Haim Bermano, Gal Chechik, and Daniel Cohen- or. An image is worth one word: Personalizing text-to- image generation using textual inversion. In The Eleventh International Conference on Learning Representations, 2023. 1, 3, 8

  7. [15]

    Interpreting clip’s image representation via text-based de- composition

    Yossi Gandelsman, Alexei A Efros, and Jacob Steinhardt. Interpreting clip’s image representation via text-based de- composition. In The Twelfth International Conference on Learning Representations, 2024. 5

  8. [16]

    Photoswap: Personalized subject swapping in images

    Jing Gu, Yilin Wang, Nanxuan Zhao, Tsu-Jui Fu, Wei Xiong, Qing Liu, Zhifei Zhang, He Zhang, Jianming Zhang, HyunJoon Jung, et al. Photoswap: Personalized subject swapping in images. Advances in Neural Information Processing Systems, 36, 2024. 2, 3 10

  9. [17]

    Swapanything: Enabling arbitrary object swapping in personalized image editing

    Jing Gu, Nanxuan Zhao, Wei Xiong, Qing Liu, Zhifei Zhang, He Zhang, Jianming Zhang, HyunJoon Jung, Yilin Wang, and Xin Eric Wang. Swapanything: Enabling arbitrary object swapping in personalized image editing. In European Conference on Computer Vision, pages 402–418, 2024. 2, 3, 5, 7

  10. [18]

    Prompt-to-prompt im- age editing with cross-attention control

    Amir Hertz, Ron Mokady, Jay Tenenbaum, Kfir Aberman, Yael Pritch, and Daniel Cohen-or. Prompt-to-prompt im- age editing with cross-attention control. In The Eleventh International Conference on Learning Representations, 2023. 8

  11. [19]

    Style aligned image generation via shared attention

    Amir Hertz, Andrey V oynov, Shlomi Fruchter, and Daniel Cohen-Or. Style aligned image generation via shared attention. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4775–4785,

  12. [20]

    Gans trained by a two time-scale update rule converge to a local nash equilibrium

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bern- hard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems, 30,

  13. [21]

    In-context lora for diffusion transformers

    Lianghua Huang, Wei Wang, Zhi-Fan Wu, Yupeng Shi, Huanzhang Dou, Chen Liang, Yutong Feng, Yu Liu, and Jingren Zhou. In-context lora for diffusion transformers. arXiv preprint arXiv:2410.23775, 2024. 3, 5

  14. [22]

    Openclip, 2021

    Gabriel Ilharco, Mitchell Wortsman, Ross Wightman, Cade Gordon, Nicholas Carlini, Rohan Taori, Achal Dave, Vaishaal Shankar, Hongseok Namkoong, John Miller, Hannaneh Hajishirzi, Ali Farhadi, and Ludwig Schmidt. Openclip, 2021. 7

  15. [23]

    Segment anything

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C Berg, Wan-Yen Lo, et al. Segment anything. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 4015–4026, 2023. 4

  16. [24]

    Tuning-free image customization with image and text guidance

    Pengzhi Li, Qiang Nie, Ying Chen, Xi Jiang, Kai Wu, Yuhuan Lin, Yong Liu, Jinlong Peng, Chengjie Wang, and Feng Zheng. Tuning-free image customization with image and text guidance. In European Conference on Computer Vision, pages 233–250. Springer, 2024. 2, 3, 5

  17. [25]

    Dreamedit: Subject-driven image editing

    Tianle Li, Max Ku, Cong Wei, and Wenhu Chen. Dreamedit: Subject-driven image editing. Transactions on Machine Learning Research, 2023. 2, 3, 5, 7

  18. [26]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceeding...

  19. [27]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Qing Jiang, Chunyuan Li, Jianwei Yang, Hang Su, et al. Grounding dino: Marrying dino with grounded pre-training for open-set object detection. In European Conference on Computer Vision, pages 38–55. Springer,

  20. [28]

    Tf- icon: Diffusion-based training-free cross-domain image composition

    Shilin Lu, Yanzhu Liu, and Adams Wai-Kin Kong. Tf- icon: Diffusion-based training-free cross-domain image composition. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2294–2305, 2023. 3, 5, 7

  21. [29]

    Rethinking the role of demonstrations: What makes in- context learning work? In EMNLP, 2022

    Sewon Min, Xinxi Lyu, Ari Holtzman, Mikel Artetxe, Mike Lewis, Hannaneh Hajishirzi, and Luke Zettlemoyer. Rethinking the role of demonstrations: What makes in- context learning work? In EMNLP, 2022. 2

  22. [30]

    Context diffusion: In-context aware image generation

    Ivona Najdenkoska, Animesh Sinha, Abhimanyu Dubey, Dhruv Mahajan, Vignesh Ramanathan, and Filip Radenovic. Context diffusion: In-context aware image generation. In European Conference on Computer Vision, pages 375–391. Springer, 2024. 2

  23. [31]

    Scalable diffusion mod- els with transformers

    William Peebles and Saining Xie. Scalable diffusion mod- els with transformers. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 4195– 4205, 2023. 1, 2

  24. [32]

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

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas M¨uller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion models for high-resolution image synthesis. In The Twelfth International Conference on Learning Representations, 2024. 1

  25. [33]

    Language models are unsupervised multitask learners

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9, 2019. 2

  26. [34]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022. 1

  27. [35]

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

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. 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, pages 22500...

  28. [36]

    Large-scale text-to-image model with inpainting is a zero-shot subject-driven image generator

    Chaehun Shin, Jooyoung Choi, Heeseung Kim, and Sungroh Yoon. Large-scale text-to-image model with inpainting is a zero-shot subject-driven image generator. arXiv preprint arXiv:2411.15466, 2024. 3, 5

  29. [37]

    Insert anything: Image insertion via in-context editing in dit

    Wensong Song, Hong Jiang, Zongxing Yang, Ruijie Quan, and Yi Yang. Insert anything: Image insertion via in-context editing in dit. arXiv preprint arXiv:2504.15009, 2025. 3

  30. [38]

    Objectstitch: Object compositing with diffusion model

    Yizhi Song, Zhifei Zhang, Zhe Lin, Scott Cohen, Brian Price, Jianming Zhang, Soo Ye Kim, and Daniel Aliaga. Objectstitch: Object compositing with diffusion model. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18310–18319, 2023. 2, 3

  31. [39]

    Imprint: Generative object compositing by learning identity-preserving representation

    Yizhi Song, Zhifei Zhang, Zhe Lin, Scott Cohen, Brian Price, Jianming Zhang, Soo Ye Kim, He Zhang, Wei Xiong, and Daniel Aliaga. Imprint: Generative object compositing by learning identity-preserving representation. In Proceedings of the IEEE/CVF Conference on Computer Vision ...

  32. [40]

    Training-free consistent 11 text-to-image generation

    Yoad Tewel, Omri Kaduri, Rinon Gal, Yoni Kasten, Lior Wolf, Gal Chechik, and Yuval Atzmon. Training-free consistent 11 text-to-image generation. ACM Transactions on Graphics (TOG), 43(4):1–18, 2024. 3

  33. [41]

    Llama: Open and efficient foundation language models.arXiv preprint arXiv:2302.13971, 2023

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timoth´ee Lacroix, Baptiste Rozi`ere, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models.arXiv preprint arXiv:2302.13971, 2023. 2

  34. [42]

    Images speak in images: A generalist painter for in-context visual learning

    Xinlong Wang, Wen Wang, Yue Cao, Chunhua Shen, and Tiejun Huang. Images speak in images: A generalist painter for in-context visual learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6830–6839, 2023. 2

  35. [43]

    Primecomposer: Faster progressively combined diffusion for image composition with attention steering

    Yibin Wang, Weizhong Zhang, Jianwei Zheng, and Cheng Jin. Primecomposer: Faster progressively combined diffusion for image composition with attention steering. In Proceedings of the 32nd ACM International Conference on Multimedia, pages 10824–10832, 2024. 2, 3, 5

  36. [44]

    In- context learning unlocked for diffusion models

    Zhendong Wang, Yifan Jiang, Yadong Lu, Pengcheng He, Weizhu Chen, Zhangyang Wang, Mingyuan Zhou, et al. In- context learning unlocked for diffusion models. Advances in Neural Information Processing Systems, 36:8542–8562,

  37. [45]

    Emergent abilities of large language models

    Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, et al. Emergent abilities of large language models. Transactions on Machine Learning Research, 2022. 2

  38. [46]

    Objectmate: A recurrence prior for object insertion and subject-driven generation

    Daniel Winter, Asaf Shul, Matan Cohen, Dana Berman, Yael Pritch, Alex Rav-Acha, and Yedid Hoshen. Objectmate: A recurrence prior for object insertion and subject-driven generation. arXiv preprint arXiv:2412.08645, 2024. 3

  39. [47]

    Head- router: A training-free image editing framework for mm- dits by adaptively routing attention heads

    Yu Xu, Fan Tang, Juan Cao, Yuxin Zhang, Xiaoyu Kong, Jintao Li, Oliver Deussen, and Tong-Yee Lee. Head- router: A training-free image editing framework for mm- dits by adaptively routing attention heads. arXiv preprint arXiv:2411.15034, 2024. 5

  40. [48]

    Paint by example: Exemplar-based image editing with diffusion models

    Binxin Yang, Shuyang Gu, Bo Zhang, Ting Zhang, Xuejin Chen, Xiaoyan Sun, Dong Chen, and Fang Wen. Paint by example: Exemplar-based image editing with diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18381– 18391, 202...

  41. [49]

    Omnipaint: Mastering object-oriented editing via disentangled insertion-removal inpainting

    Yongsheng Yu, Ziyun Zeng, Haitian Zheng, and Jiebo Luo. Omnipaint: Mastering object-oriented editing via disentangled insertion-removal inpainting. arXiv preprint arXiv:2503.08677, 2025. 3

  42. [50]

    Can mllms perform text-to-image in- context learning? arXiv preprint arXiv:2402.01293, 2024

    Yuchen Zeng, Wonjun Kang, Yicong Chen, Hyung Il Koo, and Kangwook Lee. Can mllms perform text-to-image in- context learning? arXiv preprint arXiv:2402.01293, 2024. 2

  43. [51]

    Instantswap: Fast customized concept swapping across sharp shape dif- ferences

    Chenyang Zhu, Kai Li, Yue Ma, Longxiang Tang, Chengyu Fang, Chubin Chen, Qifeng Chen, and Xiu Li. Instantswap: Fast customized concept swapping across sharp shape dif- ferences. In The Thirteenth International Conference on Learning Representations, 2024. 3 12 Supplementary Ma...

  44. [2024]

    Accessed: 2024-10-07. 2, 4

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.