Detail Continuation over a Trustworthy Coarse Scale for Autoregressive Super-Resolution

2026-08-03

AI Summary

This work addresses the hallucination problem in generative super-resolution, where synthesized high-resolution details often deviate from the low-resolution input. The authors propose a novel k-to-N detail continuation mechanism that explicitly disentangles reliable coarse-scale structure from uncertain fine-scale details within a visual autoregressive framework. Specifically, a trustworthy coarse representation is directly derived from the low-resolution input, while only the finer scales from level k to N are generated via conditional autoregression. Built upon the Visual Autoregressive (VAR) model, this approach integrates multi-scale modeling with scale-aware conditional generation to effectively mitigate error accumulation inherent in long-range autoregressive synthesis. Experiments demonstrate that the method achieves competitive performance on standard super-resolution metrics while significantly outperforming existing approaches on hallucination-aware evaluation measures, thereby enhancing structural consistency and reconstruction reliability.

Abstract

Hallucination remains a persistent challenge in generative super-resolution (GSR), where reconstructed results may contain visually plausible yet weakly supported content, structural deviations, or unnatural textures with respect to the low-resolution (LR) input. Existing GSR methods have extensively explored the trade-off between perceptual realism and reconstruction fidelity, but the division between preserving reliable coarse-scale information and restoring more uncertain fine details is often handled implicitly within the overall restoration process. Visual autoregressive (VAR) modeling provides a natural opportunity to revisit this issue, as its coarse-to-fine next-scale prediction offers an explicit scale-wise generation interface. However, existing VAR-based SR methods still inherit the original full 1-to-$N$ autoregressive generation path, even though, for super-resolution, coarse-scale information in LR is often relatively more reliable, while long autoregressive chains may accumulate prediction errors. Motivated by these observations, we propose \textbf{K2N}, which reformulates VAR-based SR from full-path generation into a $k$-to-$N$ detail continuation process. Specifically, early coarse-scale states are established directly from LR, while only the remaining finer scales are restored autoregressively. Experimental results show that K2N remains competitive with the VARSR baseline on standard SR metrics, while exhibiting clearer advantages on hallucination-focused evaluation. These findings suggest that explicitly rethinking the generation path in a scale-wise manner can be a promising direction for improving the reliability of generative super-resolution. Our code will be released soon at https://github.com/BRL-SYSU/K2NSR.

Problem

• hallucination
• generative super-resolution
• autoregressive modeling
• scale-wise generation
• reconstruction fidelity

Innovation

• autoregressive super-resolution
• hallucination reduction
• coarse-to-fine generation
• detail continuation
• visual autoregressive modeling