CoArena

Description

Builds benchmarks, leaderboards, and real-world evaluations for computer-use AI agents, helping researchers compare model performance using live tasks and blind human judgment.

Related

• Moving AtomsBuilds simulation-ready worlds from real-world scans and provides workflows and benchmarks for robotics teams using physical AI simulation…
• Standard IntelligenceDevelop pretraining models and systems that enable computer-use agents to explore, learn in-context, and perform multi-action tasks for…
• MirrorsBuilds tools that route contextual guidance into AI agent conversations, helping agent teams use trace observations, preview matched…
• MaingenBenchmark models for safely running industrial systems, provide industrial deployment research, and schedule conversations to discuss…
• Patronus AIBuilds training data, simulated environments, and evaluation software for AI systems, helping researchers and developers study model…
• Discovered MaterialsBuilds AI agents to discover materials for semiconductor chips, with work including thermal interface material simulation, synthesis,…
• TensorTestBuilds private evaluations, RL environments, and training data for AI agents on creative jobs, helping AI labs test unreleased systems and…
• TokenlessFinds flaws in reinforcement learning environments and evaluations, helping teams training models or running agent evals prevent reward…
• Standard MachinesBuilds environments and evaluations for advanced chip design, including reinforcement learning environments for long-horizon design tasks…
• Olam LabsBuilds multi-agent simulation environments and pre-deployment evaluations that help AI labs and researchers assess model safety and…
• ARC Prize FoundationBuilds human-calibrated AGI benchmarks and prize programs for open-source researchers, helping measure progress and focus work on key AI…
• CognitionBuilds tools for software creation, including an autonomous software engineering agent and code-mergeability benchmark, helping engineers…