AI Cracks Tough Math & Physics Problems
Key Questions
What math problem did the Chinese AI solve?
Peking's Rethlas/Archon AI solved a 2014 algebra conjecture proposed by an American mathematician. It accomplished this without human input in 80 hours using literature and Lean4 formal verification.
What are RL physics agents achieving?
RL physics agents rival human performance on Olympiad-level problems. They demonstrate advanced problem-solving in physics through reinforcement learning.
What do multi-user agent benchmarks reveal?
Multi-user LLM agent benchmarks highlight performance gaps in real-world deployments. Unlike single-user frameworks, they test agents handling instructions from multiple users simultaneously.
Peking Rethlas/Archon solves 2014 algebra conjecture 80hrs lit/Lean4. RL physics agents rival humans on Olympiads. Multi-user agent benchmarks highlight gaps.
Sources (2)
Updated Apr 15, 2026