Oriel — AI & technology worth noticing, every day.

World-in-Context: Interactive Video World Modeling with In-Context Conditioning

Paper · https://arxiv.org/abs/2610.33169

Interactive video world models require coordinated control over viewing perspectives, character actions, and camera motion. Existing methods typically leave view modes implicit in text captions, conflate action with camera motion, and inject controls through fragmented architectural pathways that capture interdependencies only indirectly. In this work, we present World-in-Context, an interactive world modeling framework powered by In-Context Conditioning (ICC). Rather than relying on specialized conditioning pathways, World-in-Context unifies view, action, and camera controls into explicit tokens contextualized directly within a shared self-attention space. Our formulation reveals that (i) explicit view tokens disentangle conflicting dynamics and unlock emergent view switching from single-view data, while (ii) action and camera inputs play complementary roles in governing local responsiveness and global stability. To preserve temporal continuity across execution intervals, we further introduce Interaction-Transport RoPE (IT-RoPE), parameterizing temporal attention with accumulated control transitions. On WBench, World-in-Context-5B achieves navigation performance on par with 22B EchoWM (81.5 vs. 81.6) and attains a 93.8% perspective-switching success rate, substantially outperforming existing 5B baselines (18.8%). Controlled analyses further demonstrate compounding advantages of World-in-Context under richer conditioning mixtures, favorable data efficiency, and natural applicability to causal rollouts.

First seen by Oriel on .

AI summary

World-in-Context is an interactive video world model that uses shared in-context tokens to coordinate perspective, character actions, and camera motion while preserving temporal continuity.

  • In-Context Conditioning places explicit view, action, and camera tokens within one shared self-attention space instead of specialized conditioning pathways.
  • Explicit view tokens disentangle conflicting dynamics and enable emergent perspective switching from single-view training data.
  • Interaction-Transport RoPE parameterizes temporal attention using accumulated control transitions, helping maintain continuity across execution intervals.
  • World-in-Context-5B scores 81.5 on WBench navigation versus EchoWM 22B's 81.6, while achieving 93.8% perspective-switching success versus 18.8% for existing 5B baselines.
  • Controlled analyses report stronger results with richer conditioning mixtures, favorable data efficiency, and applicability to causal rollouts.

#video world models #self-attention #in-context conditioning #interactive video

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hugging Face Papers: 34, latest observation ; +16 over observed period since

Comments

Comment count reported by the source.

  • Hugging Face Papers: 2, latest observation ; +1 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Papers2610.33169https://arxiv.org/abs/2610.33169 Read discussion