AI Paper Atlas

Research map / Video / Video generation

Arbitrary Generative Video Interpolation

Guozhen Zhang, Haiguang Wang, Chunyu Wang, Yuan Zhou, Qinglin Lu, Limin Wang

ICLR 2026 · Video › Video generation · approach: Autoregressive long video generation · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use TaRoPE instead of MLP-based timestamp injection (ablation)ArbInterp · MultiInterpBench 32xFID: 35.2 -> 33.7Improved
use TaRoPE instead of MLP-based timestamp injection (ablation)ArbInterp · MultiInterpBench 32xFVD: 454.2 -> 401.6Improved
add timestamp injection (vanilla) vs none (ablation)ArbInterp · MultiInterpBench 32xFID: 38.7 -> 35.2Improved
add appearance-motion decoupled conditioning on top of TaRoPE (ablation)ArbInterp · MultiInterpBench 32xFID: 33.7 -> 26.5Improved
add motion tokens (vs w/o motion) (ablation)ArbInterp · MultiInterpBench 32xFID: 29.8 -> 26.5Improved
add prefix-frame appearance conditioning (vs w/o appearance) (ablation)ArbInterp · MultiInterpBench 32xFID: 31.7 -> 26.5Improved
use decoupled conditioning instead of latent concatenation (ablation)ArbInterp · MultiInterpBench 32xstep time (s): 4.4 -> 2.5Improved
use decoupled conditioning instead of latent concatenation (ablation)ArbInterp · MultiInterpBench 32xFID: 27.6 -> 26.5Improved
use decoupled conditioning instead of cross-attention conditioning (ablation)ArbInterp · MultiInterpBench 32xFID: 28.4 -> 26.5Improved
use three-stage training instead of stage 3 only (ablation)ArbInterp · MultiInterpBench 32xFVD: 357.2 -> 319.9Improved
use ArbInterp over best baseline (DynamiCrafter) at 2xDynamiCrafter · MultiInterpBench 2xFID: 83.6 -> 44.9Improved
use ArbInterp over DynamiCrafter at 256x (beyond training granularity)DynamiCrafter · 256x interpolationFVD: 553.1 -> 242.3Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy