Physics simulators generate synthetic QA data for RL training that improves LLM performance on IPhO problems by 5-10 percentage points.
35 USA PHO 2019 B3 A bead of massđť‘€ slides frictionlessly along a horizontal rail
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Solving Physics Olympiad via Reinforcement Learning on Physics Simulators
Physics simulators generate synthetic QA data for RL training that improves LLM performance on IPhO problems by 5-10 percentage points.