IBM's CRAM Energy Model Trains 1,000x Faster on Sudoku
IBM researchers introduced CRAM, a learnable energy model based on Hopfield & Tank's 1980s energy approach, solving 99.996% of Sudoku-Extreme puzzles in 42ms with 100,000-215,000x speedups over TRM, while needing only 30 training examples to reach near-100% accuracy, though some critics note the benchmark is just Sudoku.
2026-11-16 ~ 2026-11-17 · 3 related posts
- CRAM energy model trains 1,000x faster, solving 99.996% of Sudoku-Extreme in 42ms — udmrzn · 2026-11-16
- CRAM hits ~100% Sudoku-Extreme accuracy with just 30 training puzzles vs 700+ for looped models — IgorCarron · 2026-11-17
- CRAM claims 100,000x+ speedup on Sudoku — 'I read the paper, it's Sudoku' — willdepue · 2026-11-17