autoregressive video generation
Think of autoregressive video generation as a model that creates videos one small chunk at a time, where each chunk is predicted based on the ones that came before it. During training, a technique called "teacher forcing" gives the model the correct previous chunks instead of its own guesses, which can accidentally let it rely on local visual similarities rather than learning about true scene changes over time. This creates a shortcut that limits the model's understanding of long-term video evolution.
The papers you shared all tackle this kind of limitation. One defines memory as persistent information that must stay accurate and influential beyond a short window, and surveys how models try to maintain it. Another introduces "Glance Forcing," which splits the denoising process between two specialized experts to reduce computational cost while keeping quality high. A third paper adds a predictive objective called Future Chunk Prediction, which forces the model to look further ahead during training, improving physical plausibility and spatiotemporal organization in the learned representations.
- Papers
- 6
- Released code
- 0
- First seen
- Oct 2026
- Latest
- Nov 2026
6 papers in the last two months, against 0 in the two before.
Start here
Three papers to read first, in the order they came out- 1Persistent memory is essential for coherent autoregressive video generation
Memory in autoregressive video generation is a five-part problem: forms, functions, operations, learning, and evaluation.
Oct 2026 - 2AI video generators learn which past frames to remember for longer clips
Long videos generated frame by frame stay coherent when the model learns which past frames to keep in memory instead of following fixed cache rules.
Nov 2026 - 3Video generators improve longer clips by revisiting earlier generation steps
Using the intermediate denoising trajectory of prior chunks as context improves autoregressive long video generation quality.
Nov 2026
How fast it is moving
Often studied with
diffusion modelA type of generative model that learns to reverse a gradual noising process to create data.4 togetherdistribution matching distillationA training method that aligns student and teacher output distributions to compress many-step generative models into efficient few-step versions.2 togetherkv cache compressionReducing the size of the key-value cache in transformers via selective retention, low-rank updates, or adaptive compression.1 togetherThe papers
Most central to this idea first, not most recent.
- Chinese Techcs.CV
Video generators improve longer clips by revisiting earlier generation steps
Using the intermediate denoising trajectory of prior chunks as context improves autoregressive long video generation quality.
Huawei Noah Research Institute · Nov 2026
- Chinese Techcs.CV
AI video generators learn which past frames to remember for longer clips
Long videos generated frame by frame stay coherent when the model learns which past frames to keep in memory instead of following fixed cache rules.
Tsinghua University, ByteDance Seed · Nov 2026
- Big Techcs.CV
Persistent memory is essential for coherent autoregressive video generation
Memory in autoregressive video generation is a five-part problem: forms, functions, operations, learning, and evaluation.
HKUST, CityUHK · Oct 2026
- Big Techcs.CV
Video translation models maintain long-term consistency by matching input regions across frames
Matching regions in input video should produce matching outputs, enabling long-range consistency in real-time video-to-video translation.
Carnegie Mellon University, Adobe Research · Nov 2026
- Chinese Techcs.CV
Future chunk prediction improves physical plausibility in autoregressive video generation
Adding Future Chunk Prediction as an auxiliary training objective improves physical plausibility in autoregressive video generation without changing inference.
Nanjing University, ARC Lab, Tencent IEG · Oct 2026
- Top Universitycs.CV
Two denoising experts make video diffusion distillation far more efficient
Glance Forcing distills bidirectional video diffusion into a few-step autoregressive model using one PF-ODE trajectory, outperforming a fully trained baseline at far lower collection cost.
Wuhan University, University of Electronic Science and Technology of China · Oct 2026
Concepts are extracted from each paper and reused across the corpus, so this page grows on its own as the desk reads.