Make it happen.

Research

Workbuddy-BenchA benchmark for evaluating coding agents on real-world work across code, web, office, and security tasks.
Youtu-AgentA framework for building, running, evaluating, and training autonomous agents.
Post-Completion-LearningCode for the paper "Post-Completion Learning for Language Models".
MCP-ZeroCode for the paper "MCP-Zero: Active Tool Discovery for Autonomous LLM Agents".
DolphinCode for the paper "Dolphin: Document Image Parsing via Heterogeneous Anchor Prompting".
NTILCode for the paper "Advancing Sequential Numerical Prediction in Autoregressive Models".
MCTBenchCode for the paper "MCTBench: Multimodal Cognition towards Text-Rich Visual Scenes Benchmark".
RepAnCode for the paper "RepAn: Enhanced Annealing through Re-parameterization".
RMI-NASCode for the paper "Neural Architecture Search With Representation Mutual Information".