OARS: Process-Aware Online Alignment for Generative Real-World Image Super-Resolution

TL;DR

OARS introduces a process-aware online alignment framework for real-world image super-resolution, leveraging a new reward model to improve perceptual quality while preserving fidelity, outperforming previous methods on benchmarks.

Contribution

The paper presents OARS, a novel online alignment method with COMPASS, a MLLM-based reward, enabling adaptive, interpretable training for super-resolution models.

Findings

Achieves state-of-the-art results on Real-ISR benchmarks.

Improves perceptual quality while maintaining fidelity.

Demonstrates effectiveness through extensive experiments and user studies.

Abstract

Aligning generative real-world image super-resolution models with human visual preference is challenging due to the perception--fidelity trade-off and diverse, unknown degradations. Prior approaches rely on offline preference optimization and static metric aggregation, which are often non-interpretable and prone to pseudo-diversity under strong conditioning. We propose OARS, a process-aware online alignment framework built on COMPASS, a MLLM-based reward that evaluates the LR to SR transition by jointly modeling fidelity preservation and perceptual gain with an input-quality-adaptive trade-off. To train COMPASS, we curate COMPASS-20K spanning synthetic and real degradations, and introduce a three-stage perceptual annotation pipeline that yields calibrated, fine-grained training labels. Guided by COMPASS, OARS performs progressive online alignment from cold-start flow matching to full-reference and finally reference-free RL via shallow LoRA optimization for on-policy exploration. Extensive experiments and user studies demonstrate consistent perceptual improvements while maintaining fidelity, achieving state-of-the-art performance on Real-ISR benchmarks.