AI Paper Atlas

Research map / Image generation / Text-to-image diffusion

Multimodal LLMs as Customized Reward Models for Text-to-Image Generation

Shijie Zhou, Ruiyi Zhang, Huaisheng Zhu, Branislav Kveton, Yufan Zhou, Jiuxiang Gu, Jian Chen, Changyou Chen

ICCV 2025 · Image generation › Text-to-image diffusion · approach: T2I evaluation and controllable adapters · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add SkipCA reward head (vs MLP) (ablation)LLaVA-Reward-Phi · MJ-Bench safetyaccuracy w/ tie: 39.7 -> 55.2Improved
add SkipCA reward head (vs MLP) (ablation)LLaVA-Reward-Phi · MJ-Bench artifact (fidelity)accuracy w/ tie: 87.3 -> 91.1Improved
add SkipCA reward head (vs MLP) (ablation)LLaVA-Reward-Phi · MJ-Bench alignmentaccuracy w/ tie: 68.2 -> 66.1Hurt
use GPM objective instead of BT (ablation)LLaVA-Reward-Phi · MJ-Bench alignmentaccuracy w/ tie: 65.2 -> 66.1Improved
add 70k hard-negative preference pairs to training (ablation)LLaVA-Reward-Phi · MJ-Bench alignmentaccuracy w/ tie: 64.5 -> 66.1Improved
use EOS-token hidden state instead of mean over tokens (ablation)LLaVA-Reward-Phi · MJ-Bench alignmentaccuracy w/ tie: 54.6 -> 66.1Improved
use LoRA instead of full fine-tuning (ablation)LLaVA-Reward-Phi · MJ-Bench artifact (fidelity)accuracy w/ tie: 86.3 -> 91.1Improved
use Qwen2.5-VL backbone instead of Phi-3.5-vision (ablation)LLaVA-Reward · MJ-Bench artifact (fidelity)accuracy w/ tie: 91.1 -> 94.3Improved
use LLaVA-NeXT 13.4B instead of Phi-3.5-vision (ablation)LLaVA-Reward · MJ-Bench artifact (fidelity)accuracy w/ tie: 91.1 -> 90.8No clear effect
use middle layer (21) hidden state instead of last layer (32) (ablation)LLaVA-Reward · SMID safetyF1 scoreHurt
use visual tokens after projector in SkipCA rather than deeper layers (ablation)LLaVA-Reward · MJ-Bench safetyaccuracyImproved
steer SDXL with LLaVA-Reward in FK steeringSDXL · GenEvaloverall score: 0.563 -> 0.645Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy