Our Vision
Building a comprehensive framework to evaluate and advance AI systems capable of genuine scientific discovery — from hypothesis to validation, from computation to experimentation
Capability Probe
Automatically collect and analyze existing benchmarks, conduct systematic tests, and identify missing capabilities in current AI systems.
Benchmark
Establish high-quality, scientist-aligned benchmarks to fill identified capability gaps with rigorous evaluation tasks.
Environment
Dynamic and open evaluation platform that can synthesize any research tasks for comprehensive capability assessment.
Coach
Train AI research agents to develop scientific taste, evaluate innovation and creativity, aligning AI with human scientific judgment.
End-to-End Auto-Research
Evaluating AI agents on complete research workflows: from literature review, hypothesis generation, experimental design, to result analysis and paper writing — measuring true autonomous research capability.
Dry-Wet Lab Integration
Bridging computational (dry lab) and experimental (wet lab) validation. Our benchmarks assess AI's ability to close the loop between in-silico predictions and real-world laboratory verification.
Towards Nobel-Level Breakthroughs
Moving beyond narrow benchmarks to evaluate AI's potential for groundbreaking scientific discoveries — the kind of creative, cross-disciplinary thinking that leads to transformative advances in human knowledge.