Research map / Large language models / Alignment and preferences

VTON-VLLM: Aligning Virtual Try-On Models with Human Preferences

NeurIPS 2025 · Large language models › Alignment and preferences · approach: Reward model design and preference models · 5 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Curates a dataset of synthesized virtual try-on images with multi-criterion human judgments and trains a vision LLM (VTON-VLLM) on these annotations to align VTON models with human preferences.

Problem

Diffusion-based virtual try-on models often fail to meet user expectations on visual quality and detail preservation, and existing evaluation does not reflect human perceptual judgments.

Approach

Learn a vision LLM judge on a dataset of VTON images annotated with human judgments across multiple perceptual criteria, so it can assess and align VTON outputs with human preferences.

Experimental setup

Datasets: synthesized VTON images with human annotations

Closest papers by meaning

• OpenVTON-Bench: A Large-Scale High-Resolution Benchmark for Controllable Virtual Try-On Evaluation (NeurIPS 2026 · similarity 0.90)
• MV-VTON: Multi-View Virtual Try-On with Diffusion Models (AAAI 2025 · similarity 0.87)
• RoVRM: A Robust Visual Reward Model Optimized via Auxiliary Textual Preference Data (AAAI 2025 · similarity 0.87)
• Incorporating Visual Correspondence into Diffusion Model for Virtual Try-On (ICLR 2025 · similarity 0.87)
• Diffusion Model Alignment Using Direct Preference Optimization (CVPR 2024 · similarity 0.86)
• VL-RewardBench: A Challenging Benchmark for Vision-Language Generative Reward Models (CVPR 2025 · similarity 0.86)