IndexAct: rethinking the index for agentic search, returns counts instead of passages
_reachsumit · x · 2026-10-07
A new arXiv paper proposes IndexAct, an Index-Native Corpus Interaction interface that decouples candidate-set refinement from text inspection:
- Agents construct and manipulate persistent candidate sets via lexical conditions and set operations over an inverted index, receiving reusable state references and statistics such as candidate counts rather than matching passages.
- Passages are requested separately when evidence is needed, and new clues can inform subsequent operations on retained sets.
- IndexAct beats evaluated baselines on five benchmarks spanning agentic search and multi-hop QA; on BrowseComp-Plus it achieves higher evidence coverage with a smaller average live context and maintains accuracy as the corpus grows.
- Analyses show informative refinement feedback and state reuse drive gains — shorter contexts or fewer search steps alone don't help.
More from Research
- You Can't Train a Model on a Discovery: Anomaly Detection as the Scientist's Shortlist — bravo_abad · 2026-10-07
- OpenAI publishes 372 AI-generated math proofs on GitHub, daring academia to keep up — The Decoder · 2026-10-07
- Slides released: 'Ultimate guide to multi-harness RL' talk from Kernel Panic Madrid — SergioPaniego · 2026-10-07
- AI Math Skeptics Pushed Back On—'Give It 2 More Months,' Researcher Quips — AvivTamar1 · 2026-10-07
- New proof bounds π's irrationality exponent at 6.0446, fully formalized in Lean 4 — Michael_D_Moor · 2026-10-07
- After 3-sum hits n^1.999, complexity theorists debate whether math is 'all square packing' — thomasahle · 2026-10-07