LENS: In-Context Search via Latent Evidence Exploration over Dynamic Raw Documents

2026-08-17

AI Summary

This study addresses the challenges of high indexing costs, rigid granularity, and update latency in dynamic document retrieval by proposing LENS, an index-free framework. LENS formulates evidence localization as a latent space exploration task under budget constraints, employing multi-strategy candidate generation and LLM-driven Bayesian belief updates to achieve query-adaptive evidence search with self-organizing cluster reuse. Experiments demonstrate that LENS attains 62.4% exact match accuracy and 84.8% evidence recall without preprocessing. On a Wikipedia subset, it achieves answer quality comparable to ReAct while improving evidence support rate to 84.0%, significantly outperforming existing baselines. These results confirm that LENS effectively balances dynamic adaptability with precise attribution, offering a robust solution for real-time retrieval-augmented generation in evolving corpora.

Abstract

LLM agents increasingly answer questions over dynamic raw-document collections, where files may change before preprocessing, and relevant evidence (spans, sections, pages, or tables) is query-dependent. Existing retrieval-augmented approaches pre-materialize evidence via fixed chunking, embeddings, or persistent indexes: effective for lookup, yet costly, stale-prone, and committed to a granularity before the query is known. We formulate in-context search as Budgeted Evidence Localization over a latent evidence space induced by dynamic raw documents and propose LENS (Latent Evidence Exploration and Search), an index-free framework. Instead of pre-materializing the evidence space, LENS maintains a query-conditioned belief over candidate units, iteratively selecting candidates via complementary lexical, local, and exploratory proposal policies, updating the belief via an LLM relevance oracle, and narrowing toward high-posterior regions under a controllable budget. Evidence is consolidated into compact, source-grounded regions of interest and compressed into self-organizing knowledge clusters reused across related queries. On a controlled 500-question evaluation with matched corpus snapshots, LENS reaches 62.4% exact match and 84.8% evidence recall vs. 65.2% exact match but 50.4% evidence recall for a ReAct-style baseline. Across scales, LENS gives the strongest supporting-fact localization and answer grounding. On a fixed 150-question fullwiki subset over the raw Wikipedia dump with zero indexing, LENS and ReAct are nearly tied in official answer quality (43.3% vs. 42.7% EM), with LENS grounding more answers in retrieved evidence (84.0% vs. 70.7%). A no-retrieval Closed-Book reference highlights the contribution of model memory. LENS is query-ready after corpus changes, needs no preprocessing or persistent index, and preserves source-grounded evidence localization throughout.

Problem

• Dynamic Raw Documents
• In-Context Search
• Evidence Localization
• Index-free Retrieval
• Retrieval-Augmented Generation

Innovation

• Index-free Framework
• Latent Evidence Exploration
• Budgeted Evidence Localization
• Dynamic Raw Documents
• Self-organizing Knowledge Clusters