HaiAI123

Curated Global AI Tools Directory

Industry update
25
heat index
Flat

Ant Group and Westlake XinChen open-source Duo v1 world model

11/18/2026 — 11/19, 10:22·1 sources (all)·1 reports (all)

Key takeaways

  • Ant Group's research institute and Hangzhou model company Westlake XinChen released and open-sourced Duo v1, a bidirectional world model. A…

Background & analysis

On November 17, 2026, Ant Institute of Technology and Hangzhou-based large model company Xihu Xinchen released and open-sourced Duo v1, a bidirectional world model officially described as a Reciprocal World Model (RWM). The model is built around two tightly coupled modules, Intent and Realization, that fold the intelligent agent into the world itself. This sets it apart from conventional world models, which generally simulate physical elements such as lighting, materials and motion and rely on keyboard, mouse or text input for user commands.

In scenarios generated by Duo v1, virtual agents share the same interactive space-time as the world. They continuously take in the user's audio and video input, decide on their own when to speak and when to listen, and generate matching audio and visual output in real time, displaying perception, reasoning and expression capabilities. One cited example shows a virtual agent saying it is thirsty after exercise, then reaching for and drinking from a bottle when the user hands over cold water.

In a human evaluation built on 30 real interaction sessions, Duo v1 passed 168 of 180 checks, giving a total pass rate of 93.3%. The model is now available in open-source form, and the available coverage does not report any further developments beyond that release.

AI-generated from 1 reports · updated 3 days ago

updream
24-hour heatHeat index 25
Just charted — trend is building up
24 hours agonow

Reports on this story headlines open the original

Nov 18
  1. Zhidongxi reports that Ant Group Research Institute and Hangzhou-based Westlake Xinchen have released and open-sourced Duo v1, a reciprocal world model. A virtual agent can continuously take in a user's audio and video, decide when to speak or listen, and generate matching sound and visuals in real time—for example, taking a water bottle handed over by the user and drinking. In 30 human evaluations, it passed 168 of 180 checks, a 93.3% pass rate.

    智东西AI score 70

Other stories people are talking about

How is heat calculated?About the method

Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.

This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.

Surge
Discussion rising fast
New
First report within 6 hours
Rising
Still gathering discussion

Back to AI industry updates →