Training Kimi K2.7 on 1,465 GDP.pdf tasks lifts GDPval too, study finds
echen · x · 2026-09-30
New research from echen's team investigates whether training on the underlying skills of GDP.pdf — now a widely used frontier eval — yields fundamental capability gains or just better PDF scores:
- They post-trained Kimi K2.7 on 1,465 GDP.pdf companion tasks.
- GDP.pdf scores improved substantially, from 11.0% to 24.5%.
- Crucially, full-task success on GDPval also improved, showing the gains transfer to broader real-economy task ability rather than merely fitting the eval format.
Paper and blog post are both available.
Related event: Kimi Research: Training on 1,465 PDF Tasks Doubles General Capability(2 posts)→
More from Research
- Google Research open-sources PaperOrchestra, a multi-agent framework for writing AI papers — google-research · 2026-11-16
- Fable 5.2's Puzzle Skills Look Suspiciously Good, Reciting Outdated MazeBench Solution — patience_cave · 2026-11-16
- Kind ICLR Reviewer Raises Score Despite Unresolved Concerns, Citing the Paper's Value — MarkSchmidtUBC · 2026-11-16
- Hand-editing 0.2% of weights plus 100 CPU-minutes turns an English TTS model into an 18MB Sanskrit one — kastnerkyle · 2026-11-16
- Why chirality matters for AI for Science: enantiomers behave differently in biology — madeofAjala · 2026-11-16
- RSS 2027 submissions open with FAQ clarifying two-stage review and negative-result bar — GeorgiaChal · 2026-11-16