Framework and protocol for open peer-to-peer collaboration between heterogeneous AI agents.

Implementation of DivInit for improving breadth scaling in agentic search via diverse query initialization.

A framework for fast data mixture optimization via gradient descent for large model training.

Benchmark and code for evaluating the robustness of translating informal mathematical proofs to Lean 4.

An 8B-scale, vision-first VL-interaction model for real-time proactive responsiveness.

A dual-expert VLA model for UAV-centric active perception and embodied question answering.

A geometric intervention to disentangle semantic signals and minimize hallucinations in VLMs.

Open-source coding environment that integrates GLM-5.2 for agentic workflows.

A framework for preserving background identity and motion continuity in long-horizon video diffusion.

Retrieval-Augmented Iterative Diffusion for cold-start and cross-lingual forecasting

A method to represent agent skills as LoRA adapters for token efficiency.

Curated list of resources and literature regarding medical world models in AI.

Framework for scaling LLM reasoning on large-scale dynamic graphs via multi-agent collaboration.

An agentic framework for autonomous deep learning workload migration via In-Context Learning.

Automated framework for uncovering hidden weaknesses in AI-generated system code.

A text-based evaluation suite for assessing reward hacking and safety in language-based agents.

A production-grade natively multimodal model powered by MSA.

An environment-engineered agent system for metric-driven autonomous scientific discovery.

A benchmark for diagnosing stage-wise hallucinations in medical MLLM reasoning.

Memory framework for visuomotor imitation policies in robotics using path signatures.