Denny Zhou: Least-to-Most Prompting Was an Early Agent Harness, Hitting 99% on SCAN
denny_zhou · x · 2026-09-08
Google researcher Denny Zhou reframes his 2022 least-to-most prompting work in today's language: it was an early agent harness — decompose a complex task into smaller ones, then repeatedly call an LLM to solve them one by one ("Decompose. Solve. Repeat.").
- On the compositional generalization benchmark SCAN, least-to-most prompting solved any split (including the length split) with at least 99% accuracy using just 14 exemplars, versus only 16% for chain-of-thought prompting.
- The retrospective connects 2022-era prompting research directly to today's agent orchestration ideas: multi-step decomposition and repeated LLM calls are the core of modern agent frameworks.
More from coding & agent
- Vercel Labs ships gpu-lexer: a 27.5KB WebGPU model for language-agnostic syntax highlighting — aidenybai · 2026-09-09
- Netlify adds GPT-6 Astra, Gemini 3.8 Flash, Claude Fable 5.1 to AI Gateway, fixes Agent Runner credits — thisiskp_ · 2026-09-09
- PSAISuite: swap LLMs in PowerShell by changing one -Model string — dfinke · 2026-09-09
- MoonPay's Paybox AI payment vault goes live in Claude, ChatGPT, Grok — kleffew94 · 2026-09-09
- 20-year IT veteran ships Train Tycoon game on Steam with Claude's help — maartendeblock · 2026-09-09
- GPT-6 'Astra' builds an interactive human anatomy app with ~4,000 structures — omarsar0 · 2026-09-08