Zhen Dong

Robot agents achieve 14% higher success rates and use 65% fewer tokens by composing primitives and performing local retries within an interactive code execution framework.

Language, when self-evolving and treated as a physical representation, can unlock state-of-the-art physical plausibility in video world models, even surpassing proprietary systems.

SafeHarness is presented, which equips the model with two obstacle-aware harnesses that enable it to prioritize the safety constraint, and surpasses the previous SOTA by 6.5% and 27.0%, respectively.

Restor boosts automated test case adoption from 74.1% to over 96%, slashing manual QA efforts while ensuring high stability in execution.