Microsoft's CorpusMap boosts agent QA by 6.4-11.7 points while cutting input tokens 34-57%
dair_ai · x · 2026-10-06
Microsoft and collaborators introduce CorpusMap, a method for agents searching large document collections.
How it works:
- Recurring entities are resolved in advance; each entity gets a page linking to every document mentioning it, while original documents stay in place.
- Agents follow entity links between documents instead of re-searching the same evidence.
- The map can be built without LLM calls and updated incrementally as new documents arrive.
Results: across 7 models and 3 benchmarks, answer quality improves 6.4-11.7 points while input tokens drop 34-57%, beating an LLM Wiki layer and three other navigation approaches.
More from coding & agent
- Justine: "Simple, readable code" often just means missed edge cases — cnakazawa · 2026-10-06
- Google Docs now supports Markdown natively, turning files into shared agent memory — Saboo_Shubham_ · 2026-10-06
- 232x faster kernel with Codex auto-research: a GPU Mode contest postmortem — dejavucoder · 2026-10-06
- Claude Code picked Preact on its own — a glimpse of AI-driven tech stack decisions — tristanbob · 2026-10-06
- Group-Evolving Agents: a new paradigm where the unit of agent self-improvement is a group — xwang_lk · 2026-10-06
- "Nobody Is Vibe-Coding a Database" — Users Only Care If It Works — sujingshen · 2026-10-06