KRAFTON, Inc. said its subsidiary Ludo Robotics presented Ludi 0.2, an updated robot intelligence system, at UC Berkeley on Oct. 30, describing gains in human interaction but releasing no scores for the version.
The company says the update improves how robots interpret social cues and gestures, retain context, and coordinate navigation with object manipulation in supported environments. Its announcement did not identify robot hardware or say when the system might be deployed. KRAFTON links the project to its experience creating AI characters and interactive game worlds. Kangwook Lee, KRAFTON’s chief AI officer and Ludo Robotics’ chief technology officer, said that “understanding and responding to people remains a major challenge” for robots.
Published test results concern Ludi 0.1, an earlier research system. An August paper tested it in 28 scenarios in AI2-THOR, a simulation. A fine-tuned Qwen3.5-9B model succeeded in 20 scenarios, while GPT-5.6 with an optimized prompt succeeded in 26. The figures do not measure Ludi 0.2 or trials with physical robots.
The paper describes Ludi 0.1 as combining speech, multimodal reasoning, memory, navigation and learned manipulation. It says a fine-tuned vision-language model selected actions while specialized policies handled navigation and object manipulation.
The symposium website lists a Ludo Robotics speaker as “TBA” for an Industry & Student Spotlight Talks session and includes a Ludo Robotics coffee break alongside student posters. Its program does not identify Ludi 0.2 or establish that the system was shown in a poster.
Fast Company reported that KRAFTON CEO CH Kim described Ludi as a research demonstration that was not yet reliable enough for everyday use.
KRAFTON describes Ludi 1.0 as a longer-term project to combine perception, dialogue and physical action in one system, without giving a target date. Fast Company reported that the company planned Ludi 1.0 for 2027.
