Yu, Jiaao, Zhang, Yanjian, Jiang, Haoyang, Wisniewski, Guillaume, Tomeh, Nadi, Jia, Shijie · Nov 4, 2026
TouchéWorld tech report : Towards Online Unified Camera--Hand Exploration in Generative Worlds
Markdown · JSON · Original source
Authors Jiaao Yu, Yanjian Zhang, Haoyang Jiang, Guillaume Wisniewski, Nadi Tomeh and Shijie Jia present TouchéWorld, a framework for adding hand gestures to video that users navigate with a camera. They say the model generates video from an initial image and scene-description prompt, conditioned on camera and hand controls aligned over . To train it, they assemble camera--hand data from egocentric videos, then combine spatial hand-skeleton conditioning with numerical hand-motion features. Causal adaptation and few-step distillation support chunk-level online generation, so users can adjust controls in response to generated visual feedback. The authors also introduce TouchéBench to evaluate whether hand motions are followed, camera and hand controls coordinate, interaction works online, and objects respond. The report frames these components as a way to study continuous exploration and hand-driven interaction in generated environments. The abstract describes the framework and evaluation areas, but does not give results or establish how well the system performs.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
cs.AI · Article
Source details
Original source date: Nov 4, 2026 (2026-11-04).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-10T00:23:05.263Z. First observed: 2026-11-05T05:06:24.065Z. Retrieved: 2026-11-10T00:18:20.999Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The supplied evidence is the abstract, which describes the framework and evaluation areas but provides no TouchéBench results. The abstract does not establish how well generated interactions work beyond describing the intended evaluation areas. The full report is not included in the supplied evidence.
Open question
How does TouchéWorld perform on the hand-motion following, camera--hand coordination, online interaction and object-response evaluations in TouchéBench?
Source-stated
The authors present TouchéWorld as a framework combining gesture-driven interaction with a camera-controlled video world model.
We present TouchéWorld, a framework that integrates gesture-driven interaction into a camera-controlled video world model.
Source-stated
The framework generates video from an initial image and scene-description prompt, conditioned on temporally aligned camera and hand controls.
the model generates video conditioned on temporally aligned camera and hand controls.
Source-stated
The authors construct joint camera--hand training data from heterogeneous egocentric videos using annotation completion, consistency-based filtering and temporal alignment.
We construct joint camera--hand training data from heterogeneous egocentric videos through annotation completion, consistency-based filtering, and temporal alignment.
Source-stated
The authors say causal adaptation and few-step distillation support chunk-level online generation, allowing users to change controls in response to generated visual feedback.
Causal adaptation and few-step distillation support chunk-level online generation, allowing users to update their controls based on generated visual feedback.
Source-stated
TouchéBench is introduced to evaluate hand-motion following, camera--hand coordination, online interaction and object response.
We further introduce TouchéBench to evaluate hand-motion following, camera--hand coordination, online interaction, and object response.
Escape Meridian interpretation
The abstract does not report TouchéWorld's performance on the evaluations it describes.
We further introduce TouchéBench to evaluate hand-motion following, camera--hand coordination, online interaction, and object response.
Source and contributor context
Yu, Jiaao, Zhang, Yanjian, Jiang, Haoyang, Wisniewski, Guillaume, Tomeh, Nadi, Jia, Shijie · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
World model
An AI system's internal model of how an environment behaves, used to predict what happens next or to plan.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: bbd7d316-c837-482c-8b26-9c86d5877322.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-10T00:23:05.263Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 705d7693ea5f4c9aeb16ba0b7419ab2b86f87c974abf316b940df6212f6c57a7. Source text hash: a83efa781e80620213d07138a587ec921f15736d112cae148949a55c185cdc9f.
Revision and source-check history
Version 1 · 2026-11-10T00:23:05.263Z
TouchéWorld tech report : Towards Online Unified Camera--Hand Exploration in Generative Worlds
Authors Jiaao Yu, Yanjian Zhang, Haoyang Jiang, Guillaume Wisniewski, Nadi Tomeh and Shijie Jia present TouchéWorld, a framework for adding hand gestures to video that users navigate with a camera. They say the model generates video from an initial image and scene-description prompt, conditioned on camera and hand controls aligned over . To train it, they assemble camera--hand data from egocentric videos, then combine spatial hand-skeleton conditioning with numerical hand-motion features. Causal adaptation and few-step distillation support chunk-level online generation, so users can adjust controls in response to generated visual feedback. The authors also introduce TouchéBench to evaluate whether hand motions are followed, camera and hand controls coordinate, interaction works online, and objects respond. The report frames these components as a way to study continuous exploration and hand-driven interaction in generated environments. The abstract describes the framework and evaluation areas, but does not give results or establish how well the system performs.
Original source
cs.AI · Article
Source details
Original source date: Nov 4, 2026 (2026-11-04).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-10T00:23:05.263Z. First observed: 2026-11-05T05:06:24.065Z. Retrieved: 2026-11-10T00:18:20.999Z.
Limits
The supplied evidence is the abstract, which describes the framework and evaluation areas but provides no TouchéBench results. The abstract does not establish how well generated interactions work beyond describing the intended evaluation areas. The full report is not included in the supplied evidence.
Open question
How does TouchéWorld perform on the hand-motion following, camera--hand coordination, online interaction and object-response evaluations in TouchéBench?
Source-stated
The authors present TouchéWorld as a framework combining gesture-driven interaction with a camera-controlled video world model.
We present TouchéWorld, a framework that integrates gesture-driven interaction into a camera-controlled video world model.
Source-stated
The framework generates video from an initial image and scene-description prompt, conditioned on temporally aligned camera and hand controls.
the model generates video conditioned on temporally aligned camera and hand controls.
Source-stated
The authors construct joint camera--hand training data from heterogeneous egocentric videos using annotation completion, consistency-based filtering and temporal alignment.
We construct joint camera--hand training data from heterogeneous egocentric videos through annotation completion, consistency-based filtering, and temporal alignment.
Source-stated
The authors say causal adaptation and few-step distillation support chunk-level online generation, allowing users to change controls in response to generated visual feedback.
Causal adaptation and few-step distillation support chunk-level online generation, allowing users to update their controls based on generated visual feedback.
Source-stated
TouchéBench is introduced to evaluate hand-motion following, camera--hand coordination, online interaction and object response.
We further introduce TouchéBench to evaluate hand-motion following, camera--hand coordination, online interaction, and object response.
Escape Meridian interpretation
The abstract does not report TouchéWorld's performance on the evaluations it describes.
We further introduce TouchéBench to evaluate hand-motion following, camera--hand coordination, online interaction, and object response.
Source and contributor context
Yu, Jiaao, Zhang, Yanjian, Jiang, Haoyang, Wisniewski, Guillaume, Tomeh, Nadi, Jia, Shijie · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
World model
An AI system's internal model of how an environment behaves, used to predict what happens next or to plan.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: bbd7d316-c837-482c-8b26-9c86d5877322.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-10T00:23:05.263Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 705d7693ea5f4c9aeb16ba0b7419ab2b86f87c974abf316b940df6212f6c57a7. Source text hash: a83efa781e80620213d07138a587ec921f15736d112cae148949a55c185cdc9f.
passed · 2026-11-10T00:23:05.263Z · anthropic/claude-opus-5.5
Publication: bbd7d316-c837-482c-8b26-9c86d5877322
attribution — pass: The reading names all six listed authors and attributes the framework, data construction, online generation and TouchéBench to them, using 'They say' and 'The authors' framing. It does not present source claims as established fact.
Authors:Jiaao Yu, Yanjian Zhang, Haoyang Jiang, Guillaume Wisniewski, Nadi Tomeh, Shijie Jia
summary — pass: The reading follows the abstract's structure: the framework, inputs and controls, training data, adaptation, online generation, benchmark, and framing. It explains the practical result of the online generation: users can adjust controls in response to generated video. Terms such as video world model, egocentric videos and few-step distillation are kept exactly, consistent with tooltip handling. The question asks about the gap the abstract leaves open, missing benchmark results, rather than making a leading claim. The source's modest framing as a framework for study is preserved, without hype.
Together, these components establish a framework for studying continuous exploration and hand-driven interaction within generative worlds.
date — pass: The originalPublished date of 2026-11-04 matches the arXiv v1 submission date. The feed date of Nov 5 is later and is correctly not used.
[Submitted on 4 Nov 2026]
support — pass: Each source-stated claim matches an abstract sentence verbatim. The interpretation claim that the abstract reports no performance results is accurate, because the abstract only lists evaluation areas. The reading implies no benefits, costs or reliability beyond the source, and its limitations correctly note that the full report is not included.
Causal adaptation and few-step distillation support chunk-level online generation, allowing users to update their controls based on generated visual feedback.
media — pass: The selected media is the verified direct arXiv abstract page for this paper, matching the source URL. No better source media was found.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
An AI system's internal model of how an environment behaves, used to predict what happens next or to plan.
Definition referenceTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivAn AI system's internal model of how an environment behaves, used to predict what happens next or to plan.
Definition referenceTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXiv