cs.CVAug 19, 2026

PALATE: Personalized Aesthetic Learning through Adaptive Taste Evolution for Multi-User Portrait Retouching

Authors: Jingxuan Wang, Yifan Mei, Yuxia Niu, Chaowan Jiao, Qijin Shen

Organizations: Harbin Institute of Technology · MT Lab, Meitu Inc. · Xiamen University · Central South University

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Automatic portrait retouching has advanced rapidly, yet its objective is inherently subjective: the same portrait admits multiple professionally valid results, and users disagree about which one is best. Most existing methods optimize a population-level aesthetic standard and therefore cannot capture individual taste, while fine-tuning a separate editing model for every user incurs prohibitive training, storage, and data costs. We propose PALATE, a shared reward-evolution framework that keeps the image editor fixed and instead personalizes the selection among retouched candidates of the same source portrait. PALATE decomposes the reward for each user into a global backbone shared by all users, category-level residuals shared by aesthetically similar users, and a lightweight user adapter, with anti-collapse regularizers keeping the three levels complementary.A cyclic dual-level distillation scheme first distills user-specific preferences into category rewards and then consolidates the resulting category-level knowledge into the global backbone, which is redistributed to initialize the next evolution round. In this way, the shared initialization improves progressively across rounds, enabling unseen users to be calibrated from only a few rankings. On expert-retouched candidates from PPR10K with held-out users and held-out images, PALATE attains 72.83% pairwise preference-prediction accuracy, surpassing all reward, aesthetic, and image-quality baselines, of which the strongest, PickScore, reaches 58.06%. Each new user costs only 512 bytes of user-specific parameters and millisecond-level scoring.

Figures & tables

Figure 1: Overview of PALATE. (a) During training, a source portrait and its retouched candidates are encoded into retouching features. A shared reward backbone and user residual adapters jointly form personalized teacher rewards. User-level preference knowledge is first distilled within each aesthetic category to update the corresponding category reward, after which category rewards are consolidated into an evolved global reward. The evolved rewards are redistributed to initialize the next round of category modeling and user adaptation. (b) During inference, a frozen image editor generates multiple retouched candidates for a source portrait, and a few-shot calibrated personalized reward ranks the candidates and selects the result that best matches the target user’s preference.
Figure 2: Global–category–user cyclic reward evolution in PALATE. Shared knowledge is distributed from the global reward to category rewards and user adapters, while personalized knowledge is distilled in the reverse direction, from users to categories and then to the global reward. The evolved global reward is redistributed to initialize the next evolution round.
Figure 3: Qualitative comparison of candidate-selection results. For each source portrait, three expert-retouched candidates are evaluated by PALATE, PickScore, and CLIPScore. The rightmost column shows the candidate selected by the user.
MethodPairwise Acc.Top-1MRRNDCG@3Exact RankHit@2Kendall τ\tauSpearman ρ\rhoMacro PairwiseWorst-prof. PairwiseMacro Top-1Worst-prof. Top-1
Aesthetic and quality assessment
FGAesQ38.3121.1752.2473.019.1749.92−0.234-0.234−0.262-0.26239.6035.7223.0817.33
CLIP-IQA43.7225.9256.9275.9213.4259.42−0.097-0.097−0.112-0.11243.9442.3326.2124.00
Vision–language similarity
CLIP image similarity41.1722.2554.5374.3912.6758.21−0.148-0.148−0.165-0.16541.2840.0022.9620.83
SigLIP similarity48.6735.8362.6978.7518.3865.880.0020.00048.7847.6736.4234.67
Table 1: Comparison under the same held-out users and source images. All metrics except Kendall’s τ\tau and Spearman’s ρ\rho are percentages. Macro scores average user profiles equally; worst-profile scores report the lowest-performing profile. Best results are bold.
Figure 4: Deployment modes under the user–image double-held-out protocol. Correct routing improves global-only scoring; wrong routing degrades it.
Figure 5: Few-shot user calibration under different support budgets. Safe calibration is stable; the vanilla adapter degrades.
Persistent storage
ComponentParametersStorage
Global reward309,5111.23 MiB
Five category rewards5×309,5115\times 309{,}5116.08 MiB
One user adapter128512 B
Computation time
OperationTimeSetting
Table 2: Storage and deployment cost measured on one NVIDIA A800-SXM4-80GB GPU with cached 195-dimensional input features. The shared PALATE state, consisting of one global and five category rewards, occupies 7.31 MiB. The adaptation timing is measured for the vanilla user adapter at K=10K=10 with 50 optimization steps and is reported as its mean and observed range over six test users.
VariantPairwise acc. (%)
Full PALATE72.83
Cyclic knowledge flow
w/o user-to-category distillation71.50
w/o category-to-global distillation72.72
w/o global broadcast72.03
w/o double teacher72.33
Table 3: Ablation results under the same user–image double-held-out protocol. All variants are selected using the validation set. The three blocks evaluate cyclic knowledge flow, input representations, and anti-collapse regularization, respectively. Pairwise accuracy is reported in percent.
More in this topic

Personalized Reward Modeling

See all 10 papers

Explore similar work

CardsList
May 9, 2026

RewardHarness: Self-Evolving Agentic Post-Training

Aligning generative models with human preferences remains a central challenge. While RLHF is widely adopted, it relies on large-scale preference data. This limitation is particularly evident in evaluating instruction-guided image edits, even though humans can often learn such preferences from only a few examples. We present RewardHarness, a training-free, self-evolving agentic reward framework that reframes reward modeling as context evolution rather than weight optimization. Instead of learning from large-scale annotations, RewardHarness aligns with human preferences by iteratively evolving a library of tools and skills from as few as 100 preference demonstrations. Given a source image, an edited image, and an editing instruction, an orchestrator agent selects the most relevant subset of tools and skills from the maintained library, and sub-agents invoke them to construct a reasoning chain that produces a reward score. By comparing predicted scores with ground-truth preferences and analyzing successes and failures in the reasoning process, the orchestrator automatically refines its library of tools and skills without additional human annotation. Using only 0.05% of the EditReward preference data, RewardHarness achieves 47.4% average accuracy on EditReward-Bench K=2/3/4 and GenAI-Bench, surpassing GPT-5 by 5.3 points. When used as a reward signal in FlowGRPO, RL-tuned models achieve 3.52 on ImgEdit-Bench. Code and model are available at the project page: https://rewardharness.com/

Yuxuan Zhang, Cong Wei, Penghui Du +11

University of British Columbia · Vector Institute · Kolors Team, Kuaishou Technology +5

Reward ModelingImage Editing EvaluationMultimodal Reward ModelingImage Generation Evaluation
AskPDF
Sep 28, 2026

PrefLUT: Reusable and Refinable Personalized Color Editing from Pairwise Preferences

Photographic color editing is inherently personal: the same image can appear too warm, too muted, or already satisfactory to different users. Most lookup table (LUT) and reference-guided methods target a specified appearance rather than model persistent preferences from repeated user choices. To address this gap, we introduce PrefLUT, a reusable and refinable user-preference modeling framework for deployable 3D LUTs, encoding ordered preferred/non-preferred image pairs into a lightweight Reusable User Profile that is reused across queries and refined using additional user preference pairs, without per-user optimization. A Query-Conditioned LUT Predictor combines this profile with each image to predict a LUT latent vector and edit strength. An Identity-Residual LUT Decoder and Edit-Strength Controller then produce an exportable 3D LUT. Experiments on three datasets demonstrate effective personalized editing and general-purpose enhancement. Each quantized profile requires only 260 bytes, and editing takes 1.365 ms/image on an RTX 5090 GPU. We also introduce the Preference-Conditioning Verification Protocol (PCVP), an evaluation protocol to verify whether personalized image edits depend on user preferences and the query image through controlled changes to user profiles, preference orders, pair correspondences, and query images.

Chuanzhi Xu, Langyi Chen, Chengkun Yue +5

The University of Sydney · Indiana University · The University of Hong Kong

Pairwise Preference LearningUser Preference ModelingImage Editing
AskPDF
Mar 1, 2026

BeautyGRPO: Aesthetic Alignment for Face Retouching via Dynamic Path Guidance and Fine-Grained Preference Modeling

Face retouching requires removing subtle imperfections while preserving unique facial identity features, in order to enhance overall aesthetic appeal. However, existing methods suffer from a fundamental trade-off. Supervised learning on labeled data is constrained to pixel-level label mimicry, failing to capture complex subjective human aesthetic preferences. Conversely, while online reinforcement learning (RL) excels at preference alignment, its stochastic exploration paradigm conflicts with the high-fidelity demands of face retouching and often introduces noticeable noise artifacts due to accumulated stochastic drift. To address these limitations, we propose BeautyGRPO, a reinforcement learning framework that aligns face retouching with human aesthetic preferences. We construct FRPref-10K, a fine-grained preference dataset covering five key retouching dimensions, and train a specialized reward model capable of evaluating subtle perceptual differences. To reconcile exploration and fidelity, we introduce Dynamic Path Guidance (DPG). DPG stabilizes the stochastic sampling trajectory by dynamically computing an anchor-based ODE path and replanning a guided trajectory at each sampling timestep, effectively correcting stochastic drift while maintaining controlled exploration. Extensive experiments show that BeautyGRPO outperforms both specialized face retouching methods and general image editing models, achieving superior texture quality, more accurate blemish removal, and overall results that better align with human aesthetic preferences.

Jiachen Yang, Xianhui Lin, Yi Dong +4

School of Cyber Science and Technology, Shenzhen Campus of Sun Yat-sen University, China · vivo BlueImage Lab, vivo Mobile Communication Co., Ltd., China · Guangdong Provincial Key Laboratory of Information Security Technology, China

Image EditingRL for Image EditingRL for Diffusion ModelsPreference-Based RLDiffusion Model Guidance
AskPDF