Tong Zhao

Even state-of-the-art MLLMs fail to grasp intuitive visual reasoning, scoring less than 60% on tasks humans solve effortlessly at a glance.

Continual Search is introduced, an iterative framework that nudges the judge, over successive turns, to keep searching for unresolved diagnostic evidence within multiple benchmark suites and model families, demonstrating that effective search supersedes raw model scale.

SetMIR not only boosts conversion rates by over 3% but also streamlines retrieval by cutting down query redundancy by a third.

BAG achieves superior performance by dynamically balancing computational efficiency and budget constraints, outperforming traditional caching methods in diverse scenarios.

Search-oriented rubrics redefine how we evaluate document sets, leading to a 2.6-point performance boost in deep research tasks.

Ling-2.6 and Ring-2.6 achieve unprecedented efficiency in agentic intelligence, enabling instant responses and deep reasoning at trillion-parameter scale.

Achieve more reliable and visually informative multimodal reports by enforcing factual grounding, citation fidelity, and cross-modal consistency throughout the workflow with a multi-agent system.

Forget finetuning video models for each robot: a single, action-free video world model can drive diverse robots when paired with a carefully designed inverse dynamics model.

Forget random sampling – this framework crafts targeted, multi-turn function-calling data that catapults smaller LLMs to state-of-the-art performance.