Applied Compute's Spork post-training recipe doubles chess puzzle solve-rate vs GRPO
ypatil125 · x · 2026-10-15
Applied Compute, a Lux Capital portfolio company, introduced Spork, an efficient post-training recipe that increases solve-rate on chess puzzles by 100% versus its best GRPO setup. The recipe: a GRPO warmup, further partitioning of task-specific datasets, training specialists on each partition, then consolidating them via multi-teacher on-policy distillation — like a spork, gains come from simply consolidating specialists.
Related event: Applied Compute's Spork Doubles Chess Puzzle Solve Rate over GRPO(2 posts)→
More from Models
- New ChatGPT Models Have Worse 'Claude Speak' Than Claude Itself, User Says — verrsane · 2026-10-15
- One-shot coding test: Grok 4.7 renders pixel-art animation in under 2 minutes, beating Opus 5.5's 10+ min — hullabaloo22 · 2026-10-15
- Meta launches Muse Wonder models, claiming SOTA on personal and professional agent tasks — AIatMeta · 2026-10-15
- Third-party eval featured on Jev site: 98.7% cheaper, 48x faster, 89% vs 88% accuracy vs GPT-5.6 Terra — generativist · 2026-10-15
- TypeSafe ships Jev 1.14.0 with image inputs, longer context under experimental 'Danger Zone' flags — usamawahabkhan · 2026-10-15
- Meta's Muse Wonder 1.0 Enters PINNACLE Top Ten, Beating Claude Opus 5.5 and GPT-6 — ryanshrout · 2026-10-15