TheoPhysEval: GPT-6-Astra solves research-level physics problems at 98% accuracy for under $3 each
IAmTimNguyen · x · 2026-10-15
A new arXiv benchmark, TheoPhysEval (2610.16733), presents 50 unpublished research-level theoretical physics problems with verifiable answers — each solvable by a world expert in about a day. Given realistic tools (literature search, numerical/symbolic scripting), the best agent, GPT-6-Astra in Codex, reaches 98% accuracy at under $3 per problem. Notably, accuracy drops only slightly without web access, suggesting significant memorization. Co-authored by 34 physicists including Jonathan Heckman and Suvrat Raju.
More from Research
- Microduck robot runs onboard SLAM with just 64 pixels of depth, 4.8 cm median error — chris_j_paxton · 2026-10-15
- UIUC paper explains sparse LLM post-training updates through the lens of Fisher information — tokenbender · 2026-10-15
- Shannon's 1971 recollection: von Neumann suggested 'entropy' because 'no one really knows what entropy really is' — eigenron · 2026-10-15
- Two-person team trains HobbyLM MoE in ~304 GPU-hours, 100-400x less compute than peers — msharmas · 2026-10-15
- AReA paper brings adaptive receptive-field attention to high-res diffusion transformers — Apprehensive_Sky892 · 2026-10-15
- ACCRUE: continual off-policy RL lets robots learn new tasks in 4-12 minutes with 98.3% skill retention — coallaoh · 2026-10-15