Running FunSearch on a 2B Model: One Hit After 1.7 Days
cephaloform · x · 2026-11-11
A developer shares results running FunSearch tasks on a 2B model: after 1.677 days and 22 steps, only a single positive hit on funsearchdudeney1917. They reflect that starting from full logit distillation of a 3.8 27B LoRA would have been a better approach.
Related event: 2B Model Hits FunSearch Target Once After 1.7 Days(2 posts)→
More from Research
- Nvidia's GR00T 2 learns from plain human video: 40% zero-shot on 50 unseen robot tasks — rohanpaul_ai · 2026-11-11
- Why AI can't cure cancer yet: Precigenetics pitches live-cell dynamic data for virtual cell models — iScienceLuvr · 2026-11-11
- Letting an agent self-improve its own harness daily: resumed runs 7→77, memory noise 94%→0 — AccBalanced · 2026-11-11
- The 2022 wooden mechanical neural network resurfaces, sparking garden-LLM musings — Michael_D_Moor · 2026-11-11
- Step-by-step guide for beginners on writing a scientific article — AnnaCiaunica · 2026-11-11
- FraudBench: AI agents commit fraud at 21.4% rate when acting autonomously, 3% when only advising — jacyanthis · 2026-11-11