Microsoft paper: distilled trajectory skills recover 55-100% of reasoning gains at 2.7-6x fewer tokens
rohanpaul_ai · x · 2026-09-07
- Microsoft researchers propose amortizing expensive test-time reasoning into reusable natural-language "skills" (arXiv 2608.07885)
- Reasoning modes beat non-reasoning on multi-step agentic tasks but pay a 3-6x output-token premium each episode, much of it re-deriving shared procedures
- Method: collect 35-50 past trajectories, have a coding agent extract recurring failure patterns, compile them into a compact skill injected into the non-reasoning model's system prompt
- Across ALFWorld, tau²-bench (telecom/retail) and SpreadsheetBench-Verified, skills recover 55%-100%+ of the reasoning gap for GPT-5.4-mini on held-out tasks, exceeding reasoning mode outright on two of four, with 2.7-6x fewer output tokens and zero reasoning tokens
- Surprisingly, skills distilled from non-reasoning trajectories alone stay competitive with those from paired reasoning corpora
- Framed as search: test-time reasoning is deep search re-paid per deployment; corpus distillation is wide search paid once
Related event: Microsoft Paper: Distilled Skill Cards Match Test-Time Reasoning(2 posts)→
More from coding & agent
- Hermes Desktop adds Routines: scheduled profiles turn agents into unattended workers — Teknium · 2026-09-07
- Pydantic launches HTTPX2, taking over stewardship of the stalled HTTPX HTTP client — samuelcolvin · 2026-09-07
- LangChain's Harrison Chase crowdsources wish list for an 'OpenRouter for tools' — hwchase17 · 2026-09-07
- Claude Code Rules silently broken since Aug 18 — disable CLAUDE_CODE_THRIFTY_SONIC to fix — 4310sy · 2026-09-07
- tenobrus: Human-centric code conventions may not survive superhuman coding models — ricklamers · 2026-09-07
- How to enable a 1M-token context window in Codex with GPT-5.6 Sol — SIGKITTEN · 2026-09-07