# Frozen History: When Temporal Losses Stop Accumulation in Learned Real-Time Supersampling

arXiv: https://arxiv.org/abs/2610.15752 · HTML: https://mlsanity.com/paper/2610.15752/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 9 of its 38 references that are in the map.
- First posted on arXiv 11 Oct 2026.

## Builds on (most cited first, 9 of 9)

- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [RepVGG: Making VGG-style ConvNets Great Again](https://mlsanity.com/paper/2101.03697/) (2021)
- [Learning Blind Video Temporal Consistency](https://mlsanity.com/paper/1808.00449/) (2018)
- [Artistic style transfer for videos](https://mlsanity.com/paper/1604.08610/) (2016)
- [ColorVideoVDP: A visual difference predictor for image, video and display distortions](https://mlsanity.com/paper/2401.11485/) (2024)
- [FuseSR: Super Resolution for Real-time Rendering through Efficient Multi-resolution Fusion](https://mlsanity.com/paper/2310.09726/) (2023)
- [Neural Super-Resolution for Real-time Rendering with Radiance Demodulation](https://mlsanity.com/paper/2308.06699/) (2023)
- [Efficient neural supersampling on a novel gaming dataset](https://mlsanity.com/paper/2308.01483/) (2023)
- [Low-latency Space-time Supersampling for Real-time Rendering](https://mlsanity.com/paper/2312.10890/) (2023)

## Abstract (arXiv)

Learned real-time supersampling accumulates jittered low-resolution samples along motion vectors. Many methods use a temporal consistency loss to reduce flicker. We show that with accurate registration, this loss can suppress accumulation: on static content, the network copies its warped history at almost no temporal cost. Camera jitter in motion vectors raises this cost and can hide the effect. We identify four registration errors that training can conceal: jitter-carrying motion vectors, one motion field for both resolution lattices, jittered depth, and an unregistered skip path. A model-free oracle shows that correct registration makes freezing two orders of magnitude cheaper on static pixels. Controlled training shows that registration reverses the temporal weight's effect on PSNR. Under exact registration, increasing the weight reduces pixel variance at the expense of accuracy and perceived quality, with a 0.7 to 0.9 dB loss on static frames at weight 0.25. This reversal recurs across three network families and two scales; the loss also reduces accuracy on unjittered STSS data. With a small temporal weight and a normalised sample accumulator, a plain registered network improves the best published QRISP PSNR by 0.34 to 0.62 dB at ×2, ×3 and ×4 upscaling to 1080p. At ×2, its static pixel standard deviation is one quarter that of DLSS 2. Without retraining, it outperforms classical temporal upsampling by at least 4.4 dB on Arm's jittered Bistro capture. Our most accurate ×4 model runs in 2.03 ms per 1080p frame on an H200. Code: https://github.com/CV-AC/Frozenhistory