cs.CVNov 3, 2026

ScoreDial: Precise Control of Visual Attributes in Image Generation via Regressor Guidance from Limited Human Ratings

Authors: Jianbo Tang, Keita Ishii, Hiroki Sakamoto, Manabu Kikuchi, Katsuma Kitamura, Chiho Semba, Kenta Nishiyama, Teppei Mori, 

Organizations: Keio University, Japan · Bridgestone Corporation, Japan

Abstract

Requested-value control is stricter than continuous steering: given cc on a human-rating scale, the generated image should attain that value from only limited ratings. We present ScoreDial, which fits a differentiable regressor RR to 100-479 human-rated images; once RR is fixed, a requested value cc is defined on its output scale, and learned target conditioning is paired with error-adaptive guidance that repeatedly reduces ∣R(x)−c∣|R(\mathbf{x})-c| during sampling. Across tire slip resistance, interior visual complexity, food desire-to-eat, and soap translucency, using one set of sampler settings, ScoreDial reaches requested values and achieves RR-MAE at most 0.013 over a range of width 0.8 and Spearman ρ=0.980ρ=0.980, the maximum attainable under our protocol. Because RR also drives sampling, we evaluate agreement with RR separately from matched sweeps, an independent auditor fitted to the same ratings on a backbone no deployed regressor uses, and blinded human judgments. MPGD-flow also reaches the requested values under RR, but ScoreDial has larger auditor displacement on every attribute, the largest among all compared methods, and a stronger MPGD-flow setting gives up target agreement without closing this gap. On every attribute, removing target conditioning leaves low RR-MAE but more than halves auditor displacement, separating precise score correction from attribute movement. Blinded rating studies with 15 raters per attribute show mean human ratings track cc on tire and soap, with linear R2=0.665R^2=0.665 and R2=0.786R^2=0.786, and in forced-choice comparisons of matched sweeps, raters chose ScoreDial's as the clearer progression in at least 84% of judgments against every baseline. Code: https://github.com/Kamijo-Gen/ScoreDial.

Figures & tables

Explore similar work

CardsList
  1. Measured Sliders: Learning Continuous Controls from Differentiable Image Measurements

    Sep 4, 2026Yijia Chen, Boyu Wei, Xuanhua YinControllable Image GenerationDiffusion-Based Image Generation

  2. Assessing Sample Quality in Conditional Generation under Compositional Shift

    Jun 8, 2026Berker Demirel, Valentino Maiorca, Marco Fumero +2Distribution ShiftConditional Generation

  3. RationalRewards: Reasoning Rewards Scale Visual Generation Both Training and Test Time

    Apr 13, 2026Haozhe Wang, Cong Wei, Weiming Ren +3Reward ModelingMultimodal Reward Modeling