Mostik's 12-PhD team tops ARC-AGI leaderboard by passing hidden states from frontier to small models
VoidAsuka · x · 2026-09-04
Mostik, a team of 12 PhDs working four months, took first place on the ARC-AGI leaderboard. Their thesis: you don't need a frontier model to generate answers, only its reasoning—so hidden states flow directly from a frontier model into a small model on your own infrastructure, no text in between. It echoes the unpublished 2024 SWAT idea (strong-to-weak ability vectors) from yuntiandeng, Woojeong Kim and Sasha Rush, which evolved into ProgramAsWeights.
Related event: Mostik Claims Top of ARC-AGI 3 with Latent-Space Model Bridging(7 posts)→
More from Models
- Scaling Law for Looped Transformers: Looping Boosts Reasoning, Not Knowledge — bookwormengr · 2026-09-04
- AI experts once pegged AGI at 2075-2100 — OpenAI's Astra shows how wrong they were — dee_hw · 2026-09-04
- Models still can't see determinants, but no-thinking time-horizons double every ~9 months — scaling01 · 2026-09-04
- GPT-6 Astra Ultra builds a Minecraft wooden house zero-shot in ~9 minutes, stairs perfect — Angaisb_ · 2026-09-04
- Benchmark saturation accelerates: ARC-AGI-3 falls in 5 months after ARC-AGI-1's 6 years — mmmbchang · 2026-09-04
- Sam Altman apologizes for messy Astra rollout, promises banked credit resets and broad API rollout soon — jxnlco · 2026-09-04