TRACES: First Benchmark for 'Discoverative AI' Measures Reasoning, Not Just Answers
iamfakhrealam · x · 2026-08-20
TRACES is the first benchmark designed to evaluate 'discoverative AI'—systems capable of solving problems without known answer keys. Unlike traditional benchmarks that test retrieval of existing answers, TRACES focuses on the reasoning process: evaluating the tools used, errors caught, and evidence gathered. The project defines 'discoverative intelligence' and offers a rubric to distinguish genuine investigation from lucky guesses. It is currently open for submissions of both solvers and hard problems.
Related event: Apodex AI Unveils TRACES, First Benchmark for Discovery AI(3 posts)→
More from Research
- TennisVAR grounds tennis tactical reasoning in stroke evidence, crushing GPT-5.5 on localization — 量子位 · 2026-08-20
- China Merchants Lab Unveils LiOS Infrastructure for Robotic Cloth Folding — 量子位 · 2026-08-20
- 2nd Edition of Advanced Data Science and Analytics with Python released with GenAI chapter — quantum_tunnel · 2026-08-20
- Interpretable ML reveals physics phase changes in material corrosion resistance — bravo_abad · 2026-08-20
- Deleting the source error doesn't fix the chat: context pollution benchmark, 72 cases — Lopsided_Scarcity979 · 2026-08-20
- Should rare classes be merged into an "Other" bucket or treated as OOD detection? — neonhexe · 2026-08-20