# Video Diffusion Alignment via Reward Gradients

Source: [Video Diffusion Alignment via Reward Gradients](<https://www.sota2.com/research/paper/68fa74161e209131cea5fbdb>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T03:02:03.292Z

Authors: Mihir Prabhudesai, Russell Mendonca, Zheyang Qin, Katerina Fragkiadaki, Deepak Pathak

Published: 2024-07-11T17:59:45.000Z

## Abstract

We have made significant progress towards building foundational video diffusion models. As these models are trained using large-scale unsupervised data, it has become crucial to adapt these models to specific downstream tasks. Adapting these models via supervised fine-tuning requires collecting target datasets of videos, which is challenging and tedious. In this work, we utilize pre-trained reward models that are learned via preferences on top of powerful vision discriminative models to adapt video diffusion models. These models contain dense gradient information with respect to generated RGB pixels, which is critical to efficient learning in complex search spaces, such as videos. We show that backpropagating gradients from these reward models to a video diffusion model can allow for compute and sample efficient alignment of the video diffusion model. We show results across a variety of reward models and video diffusion models, demonstrating that our approach can learn much more efficiently in terms of reward queries and computation than prior gradient-free approaches. Our code, model weights,and more visualization are available at https://vader-vid.github.io.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Text-to-Video Generation on HPS (V)](<https://www.sota2.com/research/sota/text-to-video-generation-on-hps-v.md>): Text-to-Video Generation; HPS (V); HPS (V) Score: 0.259.

- [Text-to-Video Generation on PickScore](<https://www.sota2.com/research/sota/text-to-video-generation-on-pickscore.md>): Text-to-Video Generation; PickScore; PickScore: 20.62.

- [Text-to-Video Generation on VBench (Quality Metrics)](<https://www.sota2.com/research/sota/text-to-video-generation-on-vbench-quality-metrics.md>): Text-to-Video Generation; VBench; Quality Score: 82.46.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2407.08737>)

[PDF](<https://arxiv.org/pdf/2407.08737.pdf>)