Pith. sign in

REVIEW 3 major objections 6 minor 2 cited by

TARA: Token-Aware LoRA for Composable Personalization in Diffusion Models

T0 review · 3 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read TARA lets independently trained LoRA modules be combined at inference time, without fusion training, by masking each module to its own rare token and aligning its spatial attention.

desk verdict TARA is a clean, plausible training-free composition method backed by a genuinely useful observation about BOS-token concentration in LoRA cross-attention, but the paper needs to clarify its LoRA layer scope and add error bars before the headline numbers can be fully trusted. read the letter →

arxiv 2508.08812 v1 pith:HYJ6A6GD submitted 2025-08-12 cs.CV

classification cs.CV
keywords personalizedtext-to-imagegenerationLoRAcompositionmulti-concepttokenfocusmaskingalignmentlossdiffusionmodelsidentitypreservationDreamBooth
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

The paper argues that when several individually trained LoRA adapters are dropped into a diffusion model at once, they fail not because the concepts are incompatible but because every adapter leans on the same beginning-of-sequence token and scatters its attention over the whole image. TARA fixes this at the source: a binary mask forces each adapter to act only on its own rare token, and a training loss nudges that token's attention toward the same image region as its class name. If correct, users can compose multiple personalized concepts from a text prompt alone, without fusion training, extra conditions, or model merging, and keep each identity intact.

What carries the argument

Token Focus Masking (TFM) is a binary mask $M_i \in \{0,1\}^{d \times n}$ shaped like the prompt embedding; it is applied to a LoRA module's Key/Value output as $f_{\Delta_i}(X) = M_i \odot (\Delta_i X)$, so only the column belonging to the rare token $x_{r_i}$ survives. The Token Alignment Loss (TAL) is $L_{\mathrm{align}} = \frac{1}{L}\sum_{l=1}^{L} \left\| W_K^{(l)} x_{\mathrm{class}} - (W_K^{(l)} + \Delta_K^{(l)}) x_r \right\|$, averaged over cross-attention layers. TFM removes token-level competition between modules; TAL makes the rare token's spatial attention follow the class token's attention, reducing feature leakage.

What would settle it

Train two TARA modules normally, then at inference disable TFM for one module while keeping its weights; if the composed image still preserves both identities, the mask is not the causal mechanism and the claimed diagnosis is wrong. Alternatively, measure per-token LoRA output magnitudes in the Query projection; if two modules' Query outputs overlap on a common token and identity loss appears, K/V-only masking is insufficient.

Watch

Extended reading notes

Core claim

The central discovery is a diagnosis and a cure. In cross-attention layers, DreamBooth LoRA modules exert most of their influence on the [BOS] token rather than on the rare token they are supposed to represent; combining modules therefore sums competing modifications into one token. Simultaneously, rare-token attention maps spread across the background and overlap with other concepts, causing feature leakage. TARA's token focus masking zeroes every LoRA output column except the rare token's column in the Key and Value projections, and its token alignment loss pulls the rare token's Key vector toward the class token's Key vector. The result is that independently trained modules can be injecte

Load-bearing premise

The method assumes that all harmful cross-module interference passes through the Key and Value projections of cross-attention; if Query, output, or feed-forward LoRA layers also conflict token-wise, masking only K and V will not stop identity loss.

Editorial extensions

If this is right

  • Independently trained TARA modules can be combined at inference simply by summing their masked outputs, with no fusion network, no weight merging, and no extra input conditions.
  • Multi-concept identity preservation improves over DreamBooth LoRA, Mix-of-Show, and orthogonal-basis LoRA on CLIP-I and DINO metrics across 2-, 3-, and 4-concept settings.
  • Single-concept generation quality stays comparable to DreamBooth LoRA, and the identity drop after adding a second module becomes much smaller.
  • The approach transfers to both SD V1.5 and SDXL backbones without changing the training recipe.
  • TFM and TAL are complementary: masking fixes identity loss driven by token competition, while alignment fixes feature leakage driven by dispersed attention.

Reading between the lines

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

  • Because TFM already isolates each module to a single token, the same masking could apply to other projection matrices (Query, output, feed-forward) if future measurements show interference there; the paper's evidence only covers Key and Value projections.
  • The class-token alignment idea suggests a testable extension to style or attribute personalization, where a style token could be aligned to a semantic descriptor token instead of a class noun.
  • TARA's inference-time composition assumes each prompt contains exactly the rare tokens used during training; an automated token-assignment mechanism would let users compose concepts without hand-crafting the prompt template.
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

3 major / 6 minor

Summary. The paper proposes TARA (Token-Aware LoRA), a method for improving multi-concept personalization in text-to-image diffusion models. TARA comprises two components: Token Focus Masking (TFM), which applies a binary mask to the output of LoRA modules in the Key/Value cross-attention projections so that each module only affects its associated rare token, and a Token Alignment Loss (TAL), which encourages the Key vector of the rare token to match that of the class token, thereby aligning spatial attention. The central claim is that this design enables training-free multi-concept composition at inference: multiple independently trained TARA LoRA modules can be injected together without additional fusion training while preserving each concept's identity. Experiments are conducted on SD V1.5 and SDXL 1.0 with DreamBooth subjects, comparing against DB-LoRA, Mix-of-Show, Custom Diffusion, and a re-implemented ROB method, with quantitative metrics CLIP-T, CLIP-I, and DINO.

Significance. If the central claim holds, TARA would be a practical and lightweight solution to a known limitation of LoRA-based personalization: composing multiple independently trained adapters without identity loss or feature leakage. The paper provides a clear and plausible mechanism (token-wise isolation in cross-attention) and includes both single-concept and multi-concept comparisons, with the DINO and CLIP-I gains over DB-LoRA being substantial. A strength is that the proposed method is modular and does not require fusion training, and the authors state that code and models will be released, which aids reproducibility. However, the empirical evaluation lacks error bars and statistical testing, and a key implementation detail about the scope of LoRA insertion is unspecified. These issues must be resolved before the central claim can be considered fully supported.

major comments (3)
  1. [§3.1, §3.3, Eq. (2), Table 3] The manuscript does not specify whether TARA inserts LoRA into all linear layers (as stated for DB-LoRA in §3.1) or only into the Key/Value projections, which is the scope where TFM is applied. If TARA trains LoRA only for K/V, then comparisons to DB-LoRA are not parameter-matched and the improvement in Tables 2 and 3 could be attributed to a smaller, less interfering adaptation rather than to the token mask. If TARA trains LoRA in all layers while masking only K/V outputs, then Query, output, and feed-forward LoRA adapters remain unmasked and could still concentrate on [BOS] or overlap, undercutting the proposed mechanism. Table 3's baseline 'LoRA' must be specified (all-layer or K/V-only), and an ablation with matched parameter counts is required to support the claim that token masking on K/V is the decisive factor.
  2. [§4.1, Tables 1-3] All quantitative claims rest on point estimates averaged over 10 generated images per prompt, but no standard deviations, confidence intervals, or significance tests are reported. For example, Table 2 shows TARA outperforming DB-LoRA by 0.09-0.14 in CLIP-I and DINO in the 2-concept setting, but without per-concept or per-prompt variance it is unclear whether these differences are consistent or driven by a few favorable cases. Please report per-subject statistics, confidence intervals, or paired significance tests to establish the robustness of the claimed improvements.
  3. [§4.1 vs. Table 2] There is a numerical inconsistency: the text states that TARA achieves CLIP-I 0.707 and DINO 0.426 in the 2-concept setting, while Table 2 reports CLIP-I 0.711 and DINO 0.431. The authors should correct this and ensure all numbers in the text and tables match exactly.
minor comments (6)
  1. [§3.4, Eq. (5)] The equation has formatting issues (stray whitespace) and L is used both as the number of cross-attention layers and in the loss name L_align. Please clarify the notation.
  2. [§3.3, Eq. (2)] The binary mask M_i is defined conceptually, but the implementation of how the rare-token column is identified during inference is not described. Is it based on the position of the special token V* in the tokenized prompt, and how is this handled when multiple rare tokens are present? Provide implementation details for reproducibility.
  3. [Table 1] The table reports 'Single' and 'Merged' results, but the merging procedure is not defined. Are LoRA weights summed, concatenated, or combined in another way? How are the TFM masks combined when merging? Clarify what '<1 s' means beyond inference time.
  4. [Figure 2(a)] The 'average magnitude of LoRA output in K and V projections' is not precisely defined. State which layers, timesteps, and texts are used, and how the averaging is performed so the analysis can be reproduced.
  5. [§3.4 and Figure 4] Because TAL explicitly optimizes the rare token's Key vector to match the class token's Key vector, the improved spatial alignment shown in Figure 4 is partially by construction. This is acceptable, but the paper should state that the primary evidence for the method's benefit is the identity-preservation metrics (CLIP-I/DINO), not the attention alignment itself.
  6. [§3.4, Eq. (6)] The TAL weighting parameter λ is fixed to 1 with no sensitivity study. Since λ controls the trade-off between the denoising objective and the auxiliary alignment objective, a small ablation over λ values (e.g., 0.1, 1, 10) would strengthen the paper.

Circularity Check

1 steps flagged · score 2.0 of 10

TARA is largely self-contained: identity-preservation claims are externally measured by CLIP-I/DINO; only the auxiliary attention-alignment effect is partly by construction.

  1. self definitional [Sec. 3.4, Eq. (4)-(5)]
    "aligning the Key vector of the rare token with that of the class token helps align their spatial attention regions... TAL encourages the Key vector of the rare token to align with that of the class token, effectively guiding the spatial attention of the rare token to align with that of the class token."

    Eq. (4) defines the spatial attention map as softmax(QK^T/sqrt(d)); for a given token, the attention distribution is a function of that token's Key vector. Eq. (5) directly minimizes ||W_K x_class - (W_K + ΔK)x_r||. When this loss is minimized, the Key vectors of the rare and class tokens coincide, so by Eq. (4) their attention maps coincide by construction. Thus the paper's claim that TAL makes the rare token's attention align with the class token is not an independent empirical finding; it is an algebraic consequence of the training objective. However, this is only a partial circularity: the paper's central identity-preservation results are evaluated with CLIP-I and DINO, which are not the same as the TAL loss, and the multi-concept gains depend on external metrics.

full rationale

The main derivation chain of TARA is not circular. TFM explicitly prevents overlapping token-wise LoRA effects by construction, but that is the proposed mechanism, not a prediction about external metrics. The identity-preservation comparisons in Tables 1-3 use CLIP-I and DINO, which are not part of the training objective (L = L_denoise + λL_align), so the headline claims have independent support. The self-citations present ([19] GLAD, [11] survey) are in related-work context and are not load-bearing for the method's validity. The only mild circularity is that the spatial-attention alignment attributed to TAL follows directly from the loss in Eq. (5) given the attention definition in Eq. (4); this is a design-consistent consequence rather than a discovered effect. The skeptic's concern about LoRA layer scope (K/V-only vs all linear layers) is an implementation-clarity issue and a potential confound for the ablation, but it is not a circularity because the paper does not use that ambiguity to redefine its evaluation. Overall circularity is low.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The method contributes a training and inference scheme rather than new physical or architectural entities. Its unknowns are hyperparameters (lambda, rank) and domain assumptions about attention behavior and rare-token uniqueness.

free parameters (2)
  • TAL weighting lambda = 1
    Chosen by hand to balance the denoising loss and the alignment loss; no sensitivity analysis is provided.
  • LoRA rank r = 8
    Fixed rank for all concepts; a standard choice but not varied, and it affects the capacity available for identity preservation.
assumptions (4)
  • standard math Cross-attention alignment of key vectors transfers to spatial attention alignment (Eq. 4 and 5).
    Attention columns are softmax(QK_i/sqrt(d)), so if two keys are equal, the attention distributions for a fixed query are identical. Invoked in Sec 3.4.
  • domain assumption Class token attention in the pretrained model is spatially localized to the concept region.
    Observed qualitatively in Figure 4, but assumed for all concepts and layers; TAL relies on it as the target. Sec 3.4.
  • domain assumption LoRA interference in multi-concept composition is dominated by cross-attention K/V projections and can be removed by restricting each module to its rare token.
    Based on Figure 2(a) visualization of K/V outputs; the paper does not quantify other projections. Sec 3.2 and 3.3.
  • domain assumption Each independently trained LoRA has a unique, identifiable rare token at inference, and the mask can be built from the prompt.
    TFM requires knowing which token corresponds to which LoRA; no mechanism is described for collision or automatic identification. Sec 3.3, Eq. 2.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TARA: Token-Aware LoRA for Composable Personalization in Diffusion Models." pith.science (2026). https://pith.science/paper/HYJ6A6GD

@misc{pith2026250808812,
  author       = {Pith},
  title        = {Pith review of: TARA: Token-Aware LoRA for Composable Personalization in Diffusion Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HYJ6A6GD}},
  note         = {Machine review of arXiv:2508.08812}
}
read the original abstract

Personalized text-to-image generation aims to synthesize novel images of a specific subject or style using only a few reference images. Recent methods based on Low-Rank Adaptation (LoRA) enable efficient single-concept customization by injecting lightweight, concept-specific adapters into pre-trained diffusion models. However, combining multiple LoRA modules for multi-concept generation often leads to identity missing and visual feature leakage. In this work, we identify two key issues behind these failures: (1) token-wise interference among different LoRA modules, and (2) spatial misalignment between the attention map of a rare token and its corresponding concept-specific region. To address these issues, we propose Token-Aware LoRA (TARA), which introduces a token mask to explicitly constrain each module to focus on its associated rare token to avoid interference, and a training objective that encourages the spatial attention of a rare token to align with its concept region. Our method enables training-free multi-concept composition by directly injecting multiple independently trained TARA modules at inference time. Experimental results demonstrate that TARA enables efficient multi-concept inference and effectively preserving the visual identity of each concept by avoiding mutual interference between LoRA modules. The code and models are available at https://github.com/YuqiPeng77/TARA.

Figures

Figures reproduced from arXiv: 2508.08812 by the authors.

Figure 1
Figure 1. Illustration of two common issues when directly apply [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the two key issues in multi-concept generation and how TARA addresses them. (a) Visualization of the average [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of the proposed TARA framework. (a) Illustration of the Token Focus Masking (TFM). Given the prompt embeddings, [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Visual examples illustrating the effect of each com [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparisons between different LoRA-based methods. All images are generated using only text prompts, without [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Sign in to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Training-Free Multi-Concept LoRA Composition with Prompt-Aware Weighting

    cs.CV 2026-06 unverdicted novelty 6.0 of 10

    Prompt-aware weighting strategies W-Switch and W-Composite improve multi-concept LoRA composition in diffusion models without training.

  2. Revisiting Efficient Semantic Segmentation: Learning Offsets for Better Spatial and Class Feature Alignment

    cs.CV 2025-08 unverdicted novelty 5.0 of 10

    OffSeg improves efficient semantic segmentation by jointly learning spatial and class offsets, adding 0.1 to 0.2M parameters to existing models.

Reference graph

Works this paper leans on

36 extracted references · 20 canonical work pages · cited by 2 Pith papers

  1. [1]

    An image is worth multiple words: Multi-attribute inversion for constrained text-to-image syn- thesis

    Aishwarya Agarwal, Srikrishna Karanam, Tripti Shukla, and Balaji Vasan Srinivasan. An image is worth multiple words: Multi-attribute inversion for constrained text-to-image syn- thesis. In 2025 IEEE/CVF Winter Conference on Applica- tions of Computer Vision (WACV), pages 6053–6062. IEEE,

  2. [2]

    Emerg- ing properties in self-supervised vision transformers

    Mathilde Caron, Hugo Touvron, Ishan Misra, Herv ´e J´egou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. Emerg- ing properties in self-supervised vision transformers. In Pro- ceedings of the IEEE/CVF international conference on com- puter vision, pages 9650–9660, 2021. 7

  3. [3]

    One-for-all: Generalized lora for parameter- efficient fine-tuning

    Arnav Chavan, Zhuang Liu, Deepak Gupta, Eric Xing, and Zhiqiang Shen. One-for-all: Generalized lora for parameter- efficient fine-tuning. arXiv preprint arXiv:2306.07967 ,

  4. [4]

    Dreambooth++: Boosting subject-driven generation via region-level references packing

    Zhongyi Fan, Zixin Yin, Gang Li, Yibing Zhan, and Heliang Zheng. Dreambooth++: Boosting subject-driven generation via region-level references packing. In Proceedings of the 32nd ACM International Conference on Multimedia , pages 11013–11021, 2024. 3

  5. [5]

    Implicit style-content separation using b-lora

    Yarden Frenkel, Yael Vinker, Ariel Shamir, and Daniel Cohen-Or. Implicit style-content separation using b-lora. In European Conference on Computer Vision, pages 181–198. Springer, 2024. 2

  6. [6]

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

    Rinon Gal, Yuval Alaluf, Yuval Atzmon, Or Patash- nik, Amit H Bermano, Gal Chechik, and Daniel Cohen- Or. An image is worth one word: Personalizing text-to- image generation using textual inversion. arXiv preprint arXiv:2208.01618, 2022. 1, 3 8

  7. [7]

    Mix-of-show: Decentralized low- rank adaptation for multi-concept customization of diffusion models

    Yuchao Gu, Xintao Wang, Jay Zhangjie Wu, Yujun Shi, Yun- peng Chen, Zihan Fan, Wuyou Xiao, Rui Zhao, Shuning Chang, Weijia Wu, et al. Mix-of-show: Decentralized low- rank adaptation for multi-concept customization of diffusion models. Advances in Neural Information Processing Sys- tems, 36:15890–15902, 2023. 2, 3, 6

  8. [8]

    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. arXiv preprint arXiv:2208.01626, 2022. 3

Show all 36 references
  1. [9]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020. 2

  2. [10]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. Lora: Low-rank adaptation of large language models. ICLR, 1(2):3, 2022. 2, 4

  3. [11]

    Diffusion model-based image editing: A survey

    Yi Huang, Jiancheng Huang, Yifan Liu, Mingfu Yan, Jiaxi Lv, Jianzhuang Liu, Wei Xiong, He Zhang, Liangliang Cao, and Shifeng Chen. Diffusion model-based image editing: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 47(6):4409–4437, 2025. 3

  4. [12]

    Identity decoupling for multi-subject personaliza- tion of text-to-image models

    Sangwon Jang, Jaehyeong Jo, Kimin Lee, and Sung Ju Hwang. Identity decoupling for multi-subject personaliza- tion of text-to-image models. Advances in Neural Informa- tion Processing Systems, 37:100895–100937, 2024. 3

  5. [13]

    Multi-concept customization of text-to-image diffusion

    Nupur Kumari, Bingliang Zhang, Richard Zhang, Eli Shechtman, and Jun-Yan Zhu. Multi-concept customization of text-to-image diffusion. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 1931–1941, 2023. 1, 3, 6

  6. [14]

    Vb-lora: Extreme pa- rameter efficient fine-tuning with vector banks

    Yang Li, Shaobo Han, and Shihao Ji. Vb-lora: Extreme pa- rameter efficient fine-tuning with vector banks. Advances in Neural Information Processing Systems, 37:16724–16751,

  7. [15]

    Ada-adapter: Fast few-shot style personlization of diffusion model with pre- trained image encoder

    Jia Liu, Changlin Li, Qirui Sun, Jiahui Ming, Chen Fang, Jue Wang, Bing Zeng, and Shuaicheng Liu. Ada-adapter: Fast few-shot style personlization of diffusion model with pre- trained image encoder. arXiv preprint arXiv:2407.05552 ,

  8. [16]

    Subject- diffusion: Open domain personalized text-to-image genera- tion without test-time fine-tuning

    Jian Ma, Junhao Liang, Chen Chen, and Haonan Lu. Subject- diffusion: Open domain personalized text-to-image genera- tion without test-time fine-tuning. In ACM SIGGRAPH 2024 Conference Papers, pages 1–12, 2024. 3

  9. [17]

    K-lora: Unlock- ing training-free fusion of any subject and style loras

    Ziheng Ouyang, Zhen Li, and Qibin Hou. K-lora: Unlock- ing training-free fusion of any subject and style loras. arXiv preprint arXiv:2502.18461, 2025. 2

  10. [18]

    Attndreambooth: To- wards text-aligned personalized text-to-image generation

    Lianyu Pang, Jian Yin, Baoquan Zhao, Feize Wu, Fu Lee Wang, Qing Li, and Xudong Mao. Attndreambooth: To- wards text-aligned personalized text-to-image generation. Advances in Neural Information Processing Systems , 37: 39869–39900, 2024. 1

  11. [19]

    Glad: Generalizable tuning for vision-language mod- els

    Yuqi Peng, Pengfei Wang, Jianzhuang Liu, and Shifeng Chen. Glad: Generalizable tuning for vision-language mod- els. arXiv preprint arXiv:2507.13089, 2025. 3

  12. [20]

    Orthogonal adaptation for modular customization of diffusion models

    Ryan Po, Guandao Yang, Kfir Aberman, and Gordon Wet- zstein. Orthogonal adaptation for modular customization of diffusion models. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 7964–7973, 2024. 2, 3, 6

  13. [21]

    Sdxl: Improving latent diffusion mod- els 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 mod- els for high-resolution image synthesis. arXiv preprint arXiv:2307.01952, 2023. 3

  14. [22]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

  15. [23]

    Exploring the limits of transfer learning with a unified text-to-text transformer

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of machine learning research, 21(140):1–67, 2020. 3

  16. [24]

    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. 3

  17. [25]

    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 2250...

  18. [26]

    Low-rank adaptation for fast text-to-image diffusion fine-tuning

    Simo Ryu. Low-rank adaptation for fast text-to-image diffusion fine-tuning. https : / / github . com / cloneofsimo/lora, 2022. 3, 4, 6

  19. [27]

    Ziplora: Any subject in any style by effectively merging loras

    Viraj Shah, Nataniel Ruiz, Forrester Cole, Erika Lu, Svet- lana Lazebnik, Yuanzhen Li, and Varun Jampani. Ziplora: Any subject in any style by effectively merging loras. In European Conference on Computer Vision, pages 422–438. Springer, 2024. 2

  20. [28]

    In- stantbooth: Personalized text-to-image generation without test-time finetuning

    Jing Shi, Wei Xiong, Zhe Lin, and Hyun Joon Jung. In- stantbooth: Personalized text-to-image generation without test-time finetuning. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 8543–8552, 2024. 3

  21. [29]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502, 2020. 2

  22. [30]

    What the daam: Interpreting stable diffu- sion using cross attention

    Raphael Tang, Linqing Liu, Akshat Pandey, Zhiying Jiang, Gefei Yang, Karun Kumar, Pontus Stenetorp, Jimmy Lin, and Ferhan Ture. What the daam: Interpreting stable diffu- sion using cross attention. arXiv preprint arXiv:2210.04885,

  23. [31]

    p+: Extended textual conditioning in text-to- image generation

    Andrey V oynov, Qinghao Chu, Daniel Cohen-Or, and Kfir Aberman. p+: Extended textual conditioning in text-to- image generation. arXiv preprint arXiv:2303.09522, 2023. 3

  24. [32]

    Sigstyle: Signature style transfer via personal- 9 ized text-to-image models

    Ye Wang, Tongyuan Bai, Xuping Xie, Zili Yi, Yilin Wang, and Rui Ma. Sigstyle: Signature style transfer via personal- 9 ized text-to-image models. In Proceedings of the AAAI Con- ference on Artificial Intelligence, pages 8051–8059, 2025. 3

  25. [33]

    Infinite-id: Identity-preserved personalization via id- semantics decoupling paradigm

    Yi Wu, Ziqiang Li, Heliang Zheng, Chaoyue Wang, and Bin Li. Infinite-id: Identity-preserved personalization via id- semantics decoupling paradigm. In European Conference on Computer Vision, pages 279–296. Springer, 2024. 3

  26. [34]

    Lori: Reducing cross-task interference in multi- task low-rank adaptation

    Juzheng Zhang, Jiacheng You, Ashwinee Panda, and Tom Goldstein. Lori: Reducing cross-task interference in multi- task low-rank adaptation. arXiv preprint arXiv:2504.07448,

  27. [35]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF international conference on computer vision, pages 3836–3847, 2023. 3

  28. [36]

    Prospect: Prompt spectrum for attribute-aware personalization of diffusion models.ACM Transactions on Graphics (TOG), 42(6):1–14, 2023

    Yuxin Zhang, Weiming Dong, Fan Tang, Nisha Huang, Haibin Huang, Chongyang Ma, Tong-Yee Lee, Oliver Deussen, and Changsheng Xu. Prospect: Prompt spectrum for attribute-aware personalization of diffusion models.ACM Transactions on Graphics (TOG), 42(6):1–14, 2023. 3 10

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.