Policy-Only Network Solves Rubik's Cube 10 Moves From Optimal on Average
haydendevs · x · 2026-11-13
A developer tested a policy network for Rubik's cube solving (not specifically trained for optimality): a perfect solver averages 18 moves, and this policy-only network is just 10 moves away. Other methods typically use test-time search; applying a 100,000-width beam search to the same weights brought the average down to 21.8 moves per solve.
Related event: Policy Network Solves Rubik's Cube Within 10 Moves of Optimal(2 posts)→
More from Models
- Startup Navi scores 73.3% on OfficeQA Pro, beating frontier labs on same model — ycombinator · 2026-11-13
- Claude account folklore in China: KYC prompts mean low weight, environment decides risk — jiayuan_jy · 2026-11-13
- OpenAI's $0.11 model beats the priciest LLMs in new Slack agent benchmark — heyshrutimishra · 2026-11-13
- Dean Ball: Open-ended long-horizon domains hit a wall near human level — verifiability is the crux — deanwball · 2026-11-13
- JevBench Splits Open-Weights Models and Hosted APIs onto Separate Leaderboards — airesearch12 · 2026-11-13
- JevBench update: top three within cost and speed caps are all open weights — airesearch12 · 2026-11-13