Google's EfficientSkill prunes agent skills in two phases, cutting output tokens up to 80%
dair_ai · x · 2026-11-17
A Google/Purdue paper introduces EfficientSkill, a two-phase framework for token-efficient multi-turn agents: first repair failures with no length penalty until target accuracy, then prune instructions and move reusable code into workspace utilities while accuracy stays close to target. Across four models on MATH-500, ALFWorld and SpreadsheetBench, output tokens fell in 11 of 12 configurations (up to 80%) while accuracy rose on every benchmark (+14.3 pts on Gemini 3.8 Flash, +17.5 on Grok 4.1 Fast), lowering total token cost with or without full-prefix KV caching.
More from coding & agent
- Codex lacks visual indicator for multi-day agent runs, user resorts to renaming tasks — chrisalbon · 2026-11-17
- Poolside's 'everything collapses into the model' thesis shapes new agent runtime Rex — sull · 2026-11-17
- FrontierCS 2 benchmark challenges AI agents to beat published CS research blind — arena · 2026-11-17
- VPS MCP Server lets AI models manage Docker containers via Portainer API — modelcontextprotocol · 2026-11-17
- Qualcomm demos an on-device decision model routing AI requests in 8.6ms on NPU — carrycooldude · 2026-11-17
- User's open letter to OpenAI: Dot agent degraded after a week, switching to Claude Duet — Danny7Guns · 2026-11-17