Paper deep dive
LENS: In-Context Search via Latent Evidence Exploration over Dynamic Raw Documents
Xingjun Wang, Gongsheng Li, Qi Fan, Yunlin Mao, Luyan Su, Yingda Chen
Intelligence
Status: succeeded | Model: Gemma-4-26B-A4B | Prompt: intel-v1 | Confidence: 91%
Last extracted: 8/23/2026, 1:36:39 AM
Summary
The paper introduces LENS (Latent Evidence Exploration and Search), an index-free framework for in-context search over dynamic raw-document collections. Unlike traditional retrieval-augmented approaches that pre-materialize evidence via fixed chunking or embeddings, LENS treats evidence localization as a budgeted sequential exploration problem over a latent evidence space. It uses a low-cost prior to narrow the search domain, then iteratively selects candidate evidence regions using lexical, local, and exploratory proposal policies, updating beliefs via an LLM-based relevance oracle. LENS consolidates evidence into source-grounded regions and self-organizing knowledge clusters. Evaluations show LENS achieves superior evidence recall and grounding compared to ReAct-style baselines, particularly in dynamic settings where documents change frequently.
Entities (6)
Relation Signals (5)
LENS → formulatesproblemas → Budgeted Evidence Localization
confidence 95% · We formulate in-context search as Budgeted Evidence Localization over a latent evidence space induced by dynamic raw documents and propose LENS...
LENS → isindexfree → true
confidence 94% · LENS is an index-free framework... needs no preprocessing or persistent index
LENS → uses → LLM Relevance Oracle
confidence 92% · updating the belief via an LLM relevance oracle, and narrowing toward high-posterior regions under a controllable budget.
LENS → outperforms → ReAct
confidence 90% · 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.
LENS → evaluatedon → Wikipedia
confidence 88% · On a fixed 150-question fullwiki subset over the raw Wikipedia dump with zero indexing...
Cypher Suggestions (0)
No Cypher suggestions yet.
Abstract
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.
Tags
Links
- Source: https://arxiv.org/abs/2608.16185v1
- Canonical: https://arxiv.org/abs/2608.16185v1
Trouble viewing inline? Open PDF directly →
Full Text
38,219 characters extracted from source content.
Expand or collapse full text
LENS: In-Context Search via Latent Evidence Exploration over Dynamic Raw Documents Xingjun Wang Gongsheng Li Qi Fan Yunlin Mao Luyan Su Yingda Chen Abstract Large language model agents increasingly need to answer questions over dynamic raw-document collections, where files may be added or updated before evidence can be preprocessed into fixed representations. Relevant evidence may appear as spans, sections, pages, or tables whose usefulness is query-dependent. Existing retrieval-augmented approaches typically materialize the evidence space before querying through fixed chunking, embeddings, or persistent indexes. While effective for lookup, such representations impose preprocessing cost, can become stale after document updates, and commit to an evidence 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 Latent Evidence Exploration and Search (LENS) as an index-free framework for this setting. Rather than pre-materializing the full evidence space, LENS maintains a query-conditioned belief over candidate evidence units, iteratively selecting candidates through complementary lexical, local, and exploratory proposal policies, updating the belief with observations from an LLM-based relevance oracle, and narrowing the search toward high-posterior evidence regions under a controllable budget. The resulting evidence is consolidated into compact, source-grounded regions of interest and further compressed into self-organizing knowledge clusters for reuse across semantically related queries. On a controlled 500-question evaluation with matched corpus snapshots, LENS achieves 62.4% exact match and 84.8% evidence recall, while a ReAct-style iterative baseline achieves 65.2% exact match but only 50.4% evidence recall. Across controlled scales, LENS consistently provides 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), while LENS grounds a larger share of answers in retrieved evidence (84.0% vs. 70.7%). A no-retrieval Closed-Book reference highlights the contribution of model memory and motivates reporting retrieval gains relative to this baseline. LENS remains query-ready immediately after corpus changes, requires no preprocessing or persistent index, and preserves source-grounded evidence localization throughout. 1 Introduction Large language model (LLM) agents increasingly operate over collections of raw documents that evolve faster than evidence can be reliably preprocessed into fixed representations. In such settings, relevant evidence is not a stable object known before the query. It may be a paragraph span, a table entry, a section, a page, or a cross-document chain whose appropriate granularity depends on both the question and the current document state. This makes document-grounded question answering different from retrieval over a static corpus of pre-segmented passages. Existing retrieval-augmented approaches typically address document question answering by materializing the evidence space before querying, through chunking, dense embeddings, summaries, persistent sparse indexes, or graph-like memory structures (7; 6; 5). These representations are effective for static lookup, especially when the corpus is stable and preprocessing cost can be amortized. However, in dynamic raw-document collections, pre-materialization introduces a trade-off: setup and update costs must be paid before querying, indexes can become stale after document changes, and fixed chunks commit to an evidence granularity before the query reveals what evidence is needed. The central difficulty is that the evidence space induced by raw documents is latent, variable-boundary, dynamic, and structured. It is latent because answer-bearing evidence exists in the documents but is not known in advance; variable-boundary because the useful evidence windows are not limited to a fixed set of chunks, which makes the space discrete but combinatorially large; dynamic because document updates change the space itself; and structured because lexical, layout, path, and historical signals induce non-uniform priors over likely evidence regions. Treating this space as a fixed finite collection of chunks can therefore obscure the actual search problem faced by an LLM agent. We formulate this setting as Budgeted Evidence Localization over a latent evidence space induced by dynamic raw documents. The goal is not merely to return the nearest precomputed passages, but to infer a compact source-grounded evidence set under token, latency, and oracle-call constraints. This formulation makes the cost of using an LLM as a relevance oracle explicit, while preserving the query-conditioned nature of evidence granularity and document freshness. To make this formulation practical, we propose Latent Evidence Exploration and Search (LENS). LENS first forms a low-cost prior over candidate evidence regions using document signals available before any expensive oracle interaction. It then performs sequential exploration: each observation from an LLM-based relevance oracle updates a belief over candidate evidence units and guides subsequent exploration toward regions with higher expected utility. Finally, LENS consolidates selected evidence into compact, source-grounded regions and organizes confirmed evidence for reuse across semantically related follow-up queries. We evaluate LENS under a dynamic raw-corpus protocol built around nested question sets and matched raw-document snapshots. This protocol measures answer quality, evidence localization, freshness under corpus growth, and query-time budget. All systems are compared under identical sampled questions and corpus boundaries, allowing paired analysis while keeping the raw-corpus setting auditable. Our contributions are as follows. First, we formulate in-context search over evolving raw-document collections as Budgeted Evidence Localization over a latent evidence space. Second, we introduce LENS, an index-free sequential exploration framework that combines low-cost priors, oracle-guided evidence refinement, budget-aware stopping, and source-grounded consolidation. Third, we design a dynamic raw-corpus evaluation protocol that jointly reports answer quality, evidence localization, freshness, budget-normalized quality, and no-retrieval reference scores. Finally, we provide mechanism evidence through source-grounding diagnostics and an ablation that isolates multi-signal prior formation from sequential exploration, rather than relying only on aggregate answer scores. 2 Background We study document collections whose contents and boundaries may change between queries. Files can be added, updated, or removed, and the evidence needed by a query may be located in spans, tables, pages, or cross-document relations. This setting differs from static open-domain retrieval because both the corpus state and the appropriate evidence granularity are query-dependent. Retrieval-augmented generation typically constructs a finite representation of the evidence space before a query arrives. Chunking, embedding indexes, sparse indexes, summary trees, and document graphs all instantiate this strategy. These representations are useful when the corpus is stable, but they commit to a fixed representation before the query is known and must be rebuilt or updated when the underlying documents change. LLMs provide strong semantic relevance judgments, but each oracle interaction consumes tokens, latency, and cost. A search method over dynamic raw documents therefore needs to trade off immediate relevance, information gain, source traceability, and budget. LENS uses this view to cast in-context search as sequential evidence localization rather than as a one-shot top-k lookup problem. 3 Budgeted Evidence Localization We consider a collection of raw documents that evolves over time. Let t=d1(t),d2(t),…,dN(t)D_t=\d_1^(t),d_2^(t),…,d_N^(t)\ (1) denote the document collection at time t. A query q arrives after the current corpus state is fixed, and the system must answer using evidence from tD_t rather than from a stale representation of t−1D_t-1. Definition 1 (Latent Evidence Space). For a dynamic raw-document collection tD_t, the latent evidence space is ℰt≜(d,s,e)∣d∈t, 0≤s<e≤|d|.E_t \(d,s,e) d _t,\ 0≤ s<e≤|d|\. (2) Each element denotes a candidate evidence window in a raw document. The space ℰtE_t is not explicitly enumerated by LENS. It is latent because the answer-bearing region is unknown before querying, variable-boundary because window boundaries may vary at span-level resolution, which makes ℰtE_t finite but combinatorially large in |d||d|, dynamic because ℰtE_t changes with tD_t, and structured because document paths, textual anchors, compiled summaries, and prior successful searches induce non-uniform beliefs over the space. Queries impose different evidence requirements. We use an intent variable I(q)∈ I(q)∈\ lookup, computation, , computation, (3) comparison, aggregation, summarize , aggregation, summarize\ to indicate whether a query can be answered by a localized span, requires multiple atomic facts, or requires a higher-level synthesis. The intent induces a set of data requirements req(q,I)=f1,…,fK,D_req(q,I)=\f_1,…,f_K\, (4) a decomposition of the query into atomic facts together with optional transformation rules that combine them. Lookup queries have K=1K=1, whereas comparison, computation, and aggregation queries have K>1K>1 and their facts may reside in different documents. Evidence localization is therefore defined per atomic fact rather than per query. Definition 2 (Per-Fact Evidence Target). For each fj∈req(q,I)f_j _req(q,I), let Zj∗∈ℰtZ_j^* _t be a minimal sufficient evidence window for fjf_j. The evidence target of the query is the collection ∗=Z1∗,…,ZK∗Z^*=\Z_1^*,…,Z_K^*\, and localization is complete only when every requirement is covered. This per-fact formulation is what makes the target well defined on multi-hop queries. A comparison between two entities needs one fact from each of two documents, so no single contiguous window is sufficient and a query-level single-window target would not exist. Defining the target per fact keeps the inference object a single window — so that beliefs and stopping rules remain tractable — while the query-level output is a set, which is also what the system returns. For a fact fjf_j, the ideal inference target is the posterior P(Zj∗∣fj,q,ℋt) P(Z_j^* f_j,q,H_t) ∝P(ℋt∣Zj∗,fj,q) P(H_t Z_j^*,f_j,q) (5) ⋅πprior(Zj∗∣fj,q,t). · _prior(Z_j^* f_j,q,D_t). where πprior _prior is a query-conditioned initial belief and ℋtH_t is the history of oracle observations. Because the prior already conditions on the query, evidence about Zj∗Z_j^* accrues through the observation likelihood. This likelihood is not available in closed form, so LENS treats an LLM as a costly relevance oracle and approximates posterior concentration through sequential observations. A budget B limits the number of oracle calls, tokens, or wall-clock time available for a query. The output is a pair (E∗,a)(E^*,a), where E∗E^* is a compact, source-grounded evidence set that covers the localized windows Z^jj≤K\ Z_j\_j≤ K and a is the synthesized answer. This distinguishes evidence localization from pure answer generation: a correct answer without traceable evidence is insufficient for the setting considered here. 4 The LENS Algorithm Figure 1: Overall LENS framework. LENS forms a query-conditioned prior over candidate evidence regions, runs a budget-constrained propose–observe–update loop, and consolidates selected regions into a compact source-grounded evidence set for answer synthesis. It never pre-materializes a persistent index over the raw-document collection. Figure 1 presents the overall framework. LENS first constructs a low-cost prior over the latent evidence space, then performs budget-constrained sequential inference in a propose–observe–update loop. Finally, it consolidates high-belief regions into a compact source-grounded evidence set for answer synthesis. This separation is central: prior formation cheaply narrows the search domain, while sequential refinement spends oracle calls only where they are informative. 4.1 Layer 1: Low-Cost Prior over Latent Evidence Directly exploring ℰtE_t is infeasible. As shown on the left of Figure 1, LENS first compresses the search space by fusing five families of signals that are available without reading the corpus through an LLM: lexical anchors, document-path structure, compiled document summaries when available, historical source-grounded evidence from prior successful searches, and lightweight corpus scans. Rather than treating such signals as independent retrieval modules, LENS uses them as approximations to a prior πprior(z∣q,t)≈∑k∈0wkπk(z∣q,t), _prior(z q,D_t)≈ _k _0w_k\, _k(z q,D_t), (6) where 0K_0 denotes a family of low-cost proposal signals. This joint prior factors, by the chain rule, into a marginal–conditional pair that separates document selection from within-document localization: πprior(z∣q,t)=πfile(dz∣q,t)πpos(sz,ez∣dz,q), _prior(z q,D_t)= _file(d_z q,D_t)\, _pos(s_z,e_z d_z,q), (7) where dzd_z is the document associated with region z. The two equations therefore describe the same object: the mixture in the first specifies how the signals are fused, and the factorization in the second specifies at which level each factor acts. Making the decomposition explicit matters because strong file-level evidence does not automatically imply precise within-document localization. Proposition 1 (Low-cost space compression). Let initC_init be the finite candidate document set induced by low-cost prior signals. Subsequent evidence exploration is restricted from ℰtE_t to search=(d,s,e)∣d∈init,0≤s<e≤|d|C_search=\(d,s,e) d _init,0≤ s<e≤|d|\. Thus, before any iterative oracle budget is consumed, LENS reduces the effective search domain from the full raw-document collection to a query-conditioned subspace. 4.2 Layer 2: Budget-Constrained Sequential Inference After prior formation, LENS enters the budgeted exploration loop shown at the center of Figure 1, which cycles through four steps while budget remains and some requirement in req(q,I)D_req(q,I) is still uncovered: (i) propose a candidate evidence region ztz_t, (i) query the LLM relevance oracle on the raw text of the region to obtain an observation oto_t, (i) update the beliefs over evidence regions, and (iv) adapt proposal weights and coverage estimates. Let ℋt=(zi,oi)i=1tH_t=\(z_i,o_i)\_i=1^t denote the observation history accumulated by this loop. Each observation is informative about every outstanding requirement, so LENS maintains one belief per fact and updates all of them from the shared history: P(Zj∗∣fj,q,ℋt) P(Z_j^* f_j,q,H_t) ∝∏i=1tP(oi∣Zj∗,zi,fj,q) _i=1^tP(o_i Z_j^*,z_i,f_j,q) (8) ⋅πprior(Zj∗∣fj,q,t). · _prior(Z_j^* f_j,q,D_t). A single oracle call therefore serves all facts at once, which is why the per-fact formulation does not multiply the oracle budget by K. The next candidate should balance exploitation and exploration: it should use the current posteriors to refine promising regions, while retaining the ability to discover evidence missed by lexical or structural priors. An ideal information-directed objective (9) selects zt+1 z_t+1 =argminz∈searchΨt(z), = _z _search _t(z), (9) Ψt(z) _t(z) =[Δt(z)]2(∗;Oz∣q,ℋt). = [ _t(z)]^2I(Z^*;O_z q,H_t). where Δt(z) _t(z) is the expected immediate relevance gap and the denominator is the expected information gain about the outstanding targets ∗Z^*. Minimizing this information ratio is principled: it targets the trade-off between immediate relevance and long-run information gain that underlies regret-optimal sequential selection, so that oracle budget is spent where it is most informative about ∗Z^*. Exact computation is intractable in raw documents, so LENS approximates this criterion with complementary proposal families: πt(z)= _t(z)= λlex(t)πlex(z)+λlocal(t)πlocal(z) _lex^(t) _lex(z)+ _local^(t) _local(z) (10) +λglobal(t)πglobal(z). + _global^(t) _global(z). Lexical proposals exploit anchors, local proposals refine around high-belief regions, and global proposals guard against semantic omissions. Treating each proposal family as an arm, LENS adapts the mixture weights λ(t)λ^(t) online from observed oracle utility, closing the propose–observe–update cycle in Figure 1 without enumerating the full latent space. 4.3 Budget-Aware Stopping The exploration loop should not continue merely because more context can be read. As depicted in Figure 1, each iteration terminates in a stopping decision: LENS exits the loop when either the remaining budget is insufficient or every requirement is localized with sufficiently concentrated belief. Following fixed-confidence best-arm identification ideas (3), a conceptual per-fact stopping statistic is GLRt(j)=min∑i≤tz≠Z^j,tlogP(oi∣Z^j,t,fj,q)P(oi∣z,fj,q). _t^(j)= _z≠ Z_j,t _i≤ t P(o_i Z_j,t,f_j,q)P(o_i z,f_j,q). (11) where Z^j,t Z_j,t is the current highest-belief region for fact fjf_j. LENS stops when minj≤KGLRt(j) _j≤ KGLR_t^(j) exceeds an intent-modulated threshold β(t,δ)γ(I)β(t,δ)\,γ(I) — that is, when the weakest requirement is resolved — or when the budget is exhausted, where γ(I)γ(I) tightens the criterion for computation and comparison intents and relaxes it for lookup. A lookup query has a single requirement and can often stop with one compact region, whereas comparison and computation queries cannot stop until each of their K facts has its own confirmed window, which is exactly the coverage condition on req(q,I)D_req(q,I). 4.4 Evidence Consolidation and Answer Synthesis Once the loop stops, the selected regions enter the consolidation-and-synthesis stage on the right of Figure 1. Consolidation merges the per-fact windows Z^jj≤K\ Z_j\_j≤ K, removes redundant or overlapping regions, expands boundaries when necessary for interpretability, and preserves source traces, yielding a compact source-grounded evidence set E∗E^*. Answer synthesis then operates on E∗E^*: for computation and comparison queries, LENS separates extraction of atomic facts from answer synthesis; for lookup-style queries, a single-stage synthesis may be sufficient. When synthesis cannot satisfy req(q,I)D_req(q,I) and budget remains, LENS triggers the self-correction path in Figure 1: a bounded step that relaxes the stopping threshold and re-enters the exploration loop with an expanded candidate set before re-synthesizing. The final output is the pair (E∗,a)(E^*,a), an answer grounded in explicit evidence regions rather than only in retrieved text snippets. 4.5 Algorithm Summary and Theoretical Properties Algorithm 1 summarizes the inference loop. The algorithm is intentionally written at the method level rather than in implementation-specific terms. Algorithm 1 LENS: Budgeted Evidence Localization 1: Input: query q, dynamic corpus tD_t, budget B 2: Build low-cost prior πprior(z∣q,t) _prior(z q,D_t) over candidate evidence regions 3: Derive requirements req(q,I)=f1,…,fKD_req(q,I)=\f_1,…,f_K\ 4: Initialize observation history ℋ0←∅H_0← and candidate subspace searchC_search 5: while budget remains and some fjf_j is uncovered do 6: Select proposal family and sample candidate region ztz_t 7: Query relevance oracle to obtain observation oto_t 8: Update per-fact beliefs P(Zj∗∣fj,q,ℋt)P(Z_j^* f_j,q,H_t) 9: Update proposal weights and requirement coverage 10: end while 11: Consolidate the per-fact windows into source-grounded evidence set E∗E^* 12: Synthesize answer a from E∗E^* and persist reusable evidence clusters when appropriate 13: Return: (E∗,a)(E^*,a) Under the abstraction above, LENS admits two analysis statements and a budget guarantee. First, low-cost priors reduce the search domain before any iterative oracle budget is spent. Second, when the oracle relevance signal is locally stable, successive observations progressively concentrate the belief over candidate evidence regions, so that additional budget yields diminishing exploration returns. The online cost of this process is bounded independently of corpus size. Proposition 2 (Bounded oracle complexity). For a loop budget of L exploration rounds, the number of LLM oracle interactions performed by LENS is bounded by c0+c1Lc_0+c_1L for small constants c0,c1c_0,c_1 that depend only on the configuration and not on the query (c0=4c_0=4, c1=2c_1=2 in our setting). The bound is independent of the number of requirements K, because one oracle observation updates all per-fact beliefs, and independent of the number of latent evidence windows induced by tD_t. These statements are analysis guides for the method rather than tight guarantees: stricter probabilistic oracle models, position-level priors, and resampling analyses remain future work. 5 Related Work Retrieval-augmented generation methods retrieve external evidence before generation and have become a standard approach for knowledge-intensive tasks (7; 6; 5). Dense, sparse, and hybrid systems are effective when a stable corpus can be preprocessed into persistent representations. Hierarchical and graph-based retrieval systems further organize documents into summaries, trees, or memory graphs (10; 4), improving reuse and traversal when the supporting structures can be built and maintained. LENS addresses a different operating point: dynamic raw-document collections where query readiness, update cost, evidence granularity, and source traceability are part of the task rather than external deployment details. Long-context language models provide another way to avoid a persistent index by placing large amounts of raw text directly into the model context (2; 8). This reduces explicit index construction, but shifts cost to query time through high token consumption and may still fail to localize the specific evidence inside long inputs. Tool-using agents can search, read, and refine their context over multiple steps (12; 1), but they often lack an explicit formulation of evidence localization under a controllable budget. LENS treats interaction as posterior-guided evidence exploration over a latent evidence space, and evaluates the resulting trade-off through answer quality, source-grounded evidence localization, freshness, and lifecycle cost. 6 Experiments 6.1 Setup We evaluate LENS on HotpotQA fullwiki (11), a multi-hop question answering benchmark where each question requires reasoning over two or more Wikipedia articles. We report results under two conditions: Controlled evaluation (DnD_n). From the validation split (7,405 questions) we draw frozen, proportionally stratified evaluation sets over ×type×difficulty strata (seed 42). Our primary controlled evaluation uses n=500n=500 questions paired with a matched corpus snapshot D500D_500 containing the gold supporting articles, context distractors, and a deterministic background pool. This is the largest statistically robust scale we evaluate and enables paired comparison across all baselines including index-dependent systems. Open-domain fullwiki. The complete raw Wikipedia dump underlying the fullwiki validation split is stored as 15,517 JSON shards without preprocessing, chunking, or indexing. We evaluate n=150n=150 fixed questions on this corpus to measure corpus-scale robustness under the zero-index constraint. Systems. We compare five arms: (1) LENS, the full algorithm with multi-signal prior, sequential exploration, budget-aware stopping, and evidence consolidation; (2) ReAct Search, a strong iterative baseline using ReAct-style (12) tool-use reasoning with the same corpus access but without LENS’s structured prior formation or budget-constrained belief updates; (3) Hybrid-RAG, a retrieval-augmented baseline combining BM25 and dense embedding retrieval with a pre-materialized index; (4) BM25-RAG, a sparse-retrieval baseline using a pre-built BM25 index; and (5) Closed-Book, a no-retrieval reference that estimates the score attributable to model parameters alone. Systems (3) and (4) require pre-materialized indexes and therefore face a lifecycle limitation under corpus changes. Hyperparameters. All systems share one chat backend (Qwen3.7, a 35B-parameter mixture-of-experts model with 3B active parameters, fixed temperature, 120 s per-call timeout) under a 300 s per-question wall-clock cap. LENS runs its DEEP configuration with a 128K query-time token budget and at most 10 candidate files admitted to evidence extraction. All runs use cold caches with no knowledge reuse. Environment. Experiments run on a single Apple M4 Pro workstation (12 cores, 48 GB RAM, macOS 15) with no local GPU; all model calls are served by a remote OpenAI-compatible endpoint, so reported latencies include network time. At most 5 questions are evaluated concurrently. Metrics. EM and F1 are the official answer metrics. Ev.Rec measures retrieval of gold supporting-fact documents, while Ground measures whether the final answer is traceable to retrieved evidence. Judge is an auxiliary semantic-equivalence diagnostic from an independent model and is not used as the primary score. Snapshot Samples Articles Evidence Background D125D_125 125 5,416 250 4,062 D250D_250 250 10,808 500 8,106 D500D_500 500 21,424 996 16,068 System (D500D_500) Ready Rebuild Index(s) Storage LENS yes no 0.0 0 ReAct yes no 0.0 0 BM25-RAG no yes 4.0 10.0MB Hybrid-RAG no yes 5.7 54.9MB Table 1: Snapshot construction and lifecycle audit. The upper block reports deterministic DnD_n corpus sizes; the lower block reports D500D_500 query readiness, rebuild requirement, index time, and index storage. 6.2 Main Results: Controlled Evaluation System EM F1 Ev.Rec Ground ReAct Search 65.2 78.9 50.4 71.8 LENS 62.4 76.9 84.8 96.8 Hybrid-RAG 38.4 51.1 80.8 92.2 BM25-RAG 28.8 42.3 71.8 95.8 Closed-Book 35.2 47.0 0.0 0.0 Table 2: Controlled evaluation on D500D_500 (%, n=500). EM/F1 are official answer metrics; Ev.Rec is supporting-fact document recall; Ground is the percentage of answers traceable to retrieved evidence. Table 2 reports the primary controlled evaluation. ReAct Search attains the highest answer score (65.2% EM, 78.9% F1), while LENS remains close on answer quality (62.4% EM, 76.9% F1) and provides substantially stronger evidence localization. LENS achieves 84.8% evidence recall and 96.8% grounded answers, compared with 50.4% and 71.8% for ReAct. This separates two evaluation dimensions: ReAct more often produces the exact answer string, whereas LENS more reliably localizes and traces the supporting evidence. Relative to Closed-Book, LENS gains 27.2 p EM and ReAct gains 30.0 p, so retrieval benefits are reported against a no-retrieval reference rather than assumed from answer correctness alone. 6.3 Open-Domain Fullwiki Results System EM F1 Ev.Rec Ground LENS 43.3 57.5 45.0 84.0 ReAct Search 42.7 57.3 50.4 70.7 Closed-Book 38.7 49.7 0.0 0.0 Table 3: Open-domain fullwiki results (%, n=150, fixed sample IDs). LENS and ReAct search the raw Wikipedia dump (15,517 shards) with zero indexing; Closed-Book reads no corpus. Table 3 reports results on the full raw Wikipedia corpus. LENS and ReAct are effectively tied on official answer quality (43.3% vs. 42.7% EM), while LENS grounds more answers in retrieved evidence (84.0% vs. 70.7%). Closed-Book reaches 38.7% EM, so fullwiki retrieval gains are modest but positive: +4.6 p for LENS and +4.0 p for ReAct. BM25-RAG and Hybrid-RAG require separate full-dump indexing and are evaluated through the controlled/lifecycle arms. 6.4 Evidence Recall Leadership System D125D_125 D250D_250 D500D_500 LENS 89.1 85.9 84.8 Hybrid-RAG 81.3 77.8 80.8 BM25-RAG 71.6 71.1 71.8 ReAct Search 46.1 47.9 50.4 Closed-Book 0.0 0.0 0.0 Table 4: Evidence recall (%) across controlled evaluation scales. LENS consistently provides the strongest supporting-fact localization, while Closed-Book has no evidence trace by design. Table 4 presents evidence recall across all controlled evaluation scales. LENS is the top evidence-localization system at every scale, with 84.8–89.1% evidence recall. Its lead over ReAct is 43.0 p on D125D_125, 38.0 p on D250D_250, and 34.4 p on D500D_500. Hybrid-RAG also retrieves substantial evidence, but its answer quality remains far below LENS and ReAct, indicating that locating candidate evidence and synthesizing the exact multi-hop answer are separable failure modes. 6.5 Corpus Staleness and Lifecycle Robustness Exact Match (%) Evidence Recall (%) System Fresh Stale Δ Fresh Stale Δ BM25-RAG 30.4 2.4 −28.0-28.0 70.1 0.0 −70.1-70.1 Hybrid-RAG 42.4 13.6 −28.8-28.8 75.1 5.5 −69.6-69.6 ReAct Search 72.0 72.8 +0.8+0.8 54.1 48.3 −5.7-5.7 LENS 69.6 67.2 −2.4-2.4 84.7 83.9 −0.8-0.8 Table 5: Staleness robustness under corpus expansion (D125→D250D_125→ D_250, Δn=125 n=125). Index-heavy systems reuse an index built on D125D_125 while answering newly added D250D_250 questions; index-free systems query the updated corpus directly. Negative gaps indicate stale degradation; the table measures freshness/lifecycle robustness rather than fresh-corpus answer quality. The stale-index arm measures what happens when an index built on D125D_125 is reused after the corpus expands to D250D_250. BM25-RAG and Hybrid-RAG are index-dependent and therefore lose most of their ability to answer newly added questions: EM drops by 28.0 and 28.8 p, and evidence recall drops by 70.1 and 69.6 p. ReAct and LENS are index-free, so they can query the expanded corpus immediately; their EM changes are small (+0.8 p and -2.4 p), and LENS retains nearly all supporting-fact recall (84.7% to 83.9%). Table 1 shows the corresponding lifecycle condition: index-heavy systems are not query-ready after corpus growth and require rebuilding, while LENS and ReAct remain query-ready. 6.6 Cost and Statistical Checks Paired McNemar tests on official EM do not show a significant LENS–ReAct answer-quality difference on D500D_500 (62.4% vs. 65.2%, p=0.1143p=0.1143) or fullwiki dev-150 (43.3% vs. 42.7%, p=1.0000p=1.0000). Query-time budget shows the expected trade-off: on D500D_500, ReAct uses 11.8K tokens per query, while LENS uses 16.5K but improves evidence recall by 34.4 p and grounding by 25.0 p. Auxiliary JudgeAcc follows EM and is not used as a primary score. 6.7 Ablation Study Configuration EM F1 Ev.Rec Ground LENS (Full) 43.3 57.5 45.0 84.0 w/o Multi-signal Prior 43.3 55.9 45.4 85.3 w/o Sequential Exploration 38.0 50.9 31.0 82.0 Table 6: Ablation study on the fullwiki evaluation (n=150 fixed sample IDs). Ev.Rec is supporting-fact document recall; Ground is the percentage of answers traceable to retrieved evidence. Table 6 isolates two LENS components on the fixed fullwiki subset. Removing sequential exploration produces the clearest degradation: EM falls from 43.3% to 38.0%, F1 from 57.5% to 50.9%, and evidence recall from 45.0% to 31.0%. Removing the multi-signal prior does not reduce EM on this subset, but it lowers F1 and changes the cost profile. 7 Discussion and Conclusions Across controlled corpora, LENS is the strongest evidence-localization system: on D500D_500 it trails ReAct by 2.8 p EM, but leads by 34.4 p evidence recall and 25.0 p grounding. On fullwiki dev-150, the two systems are effectively tied in official EM/F1 while LENS retains stronger grounding. These results support a precise claim: LENS is not an EM-dominant answer generator, but an index-free evidence localization method that makes answers more traceable to current raw sources. This distinction matters because EM is a narrow string-match measure. It penalizes acceptable paraphrases and aliases, and it can reward a correct answer that is produced from model memory rather than from retrieved evidence. The Closed-Book reference makes this visible: a non-retrieval model already reaches 35.2% EM on D500D_500 and 38.7% EM on fullwiki. We therefore treat EM/F1 as primary answer-quality metrics, but interpret them together with Ev.Rec and Ground when evaluating systems intended for auditable, source-grounded search. LENS’s advantages appear on three dimensions. First, it localizes evidence more reliably than ReAct across D125D_125, D250D_250, and D500D_500, indicating that the sequential search loop finds supporting documents rather than only plausible answers. Second, it keeps a high grounding rate, which is essential when users need to inspect the source trail. Third, the stale-index arm shows a lifecycle advantage: BM25-RAG and Hybrid-RAG lose 28.0–28.8 p EM and nearly all evidence recall when a D125D_125 index is used on new D250D_250 questions, whereas index-free systems remain query-ready over the updated corpus. This is a freshness and cost-structure advantage, not a claim that LENS always improves fresh-corpus EM. The present evidence is still bounded. HotpotQA emphasizes lookup and comparison over encyclopedic text, so richer document layouts, aggregation intents, table evidence, and warm-reuse behavior remain future work. Within this scope, LENS demonstrates competitive answer quality, substantially stronger source traceability, and immediate query readiness over changing raw documents—the operating point targeted by Budgeted Evidence Localization. References Asai et al. (2024) A. Asai, Z. Wu, Y. Wang, A. Sil, and H. Hajishirzi Self-rag: learning to retrieve, generate, and critique through self-reflection. In International conference on learning representations, p. 9112–9141. Cited by: §5. Beltagy et al. (2020) I. Beltagy, M. E. Peters, and A. Cohan Longformer: the long-document transformer. arXiv preprint arXiv:2004.05150. Cited by: §5. Garivier and Kaufmann (2016) A. Garivier and E. Kaufmann Optimal best arm identification with fixed confidence. In Conference on Learning Theory, p. 998–1027. Cited by: §4.3. Gutiérrez et al. (2024) B. J. Gutiérrez, Y. Shu, Y. Gu, M. Yasunaga, and Y. Su HippoRAG: neurobiologically inspired long-term memory for large language models. In Advances in Neural Information Processing Systems, p. 59532–59569. Cited by: §5. Izacard et al. (2023) G. Izacard, P. Lewis, M. Lomeli, L. Hosseini, F. Petroni, T. Schick, J. Dwivedi-Yu, A. Joulin, S. Riedel, and E. Grave Atlas: few-shot learning with retrieval augmented language models. Journal of Machine Learning Research 24 (251). Cited by: §1, §5. Karpukhin et al. (2020) V. Karpukhin, B. Oguz, S. Min, P. Lewis, L. Wu, S. Edunov, D. Chen, and W. Yih Dense passage retrieval for open-domain question answering. In Proceedings of the 2020 conference on empirical methods in natural language processing (EMNLP), p. 6769–6781. Cited by: §1, §5. Lewis et al. (2020) P. Lewis, E. Perez, A. Piktus, F. Petroni, V. Karpukhin, N. Goyal, H. Küttler, M. Lewis, W. Yih, T. Rocktäschel, S. Riedel, and D. Kiela Retrieval-augmented generation for knowledge-intensive nlp tasks. In Advances in Neural Information Processing Systems, p. 9459–9474. Cited by: §1, §5. Liu et al. (2024) N. F. Liu, K. Lin, J. Hewitt, A. Paranjape, M. Bevilacqua, F. Petroni, and P. Liang Lost in the middle: how language models use long contexts. Transactions of the Association for Computational Linguistics 12, p. 157–173. Cited by: §5. Russo and Van Roy (2014) D. Russo and B. Van Roy Learning to optimize via information-directed sampling. In Advances in Neural Information Processing Systems, Cited by: §4.2. Sarthi et al. (2024) P. Sarthi, S. Abdullah, A. Tuli, S. Khanna, A. Goldie, and C. Manning Raptor: recursive abstractive processing for tree-organized retrieval. In International Conference on Learning Representations, p. 32628–32649. Cited by: §5. Yang et al. (2018) Z. Yang, P. Qi, S. Zhang, Y. Bengio, W. Cohen, R. Salakhutdinov, and C. D. Manning HotpotQA: a dataset for diverse, explainable multi-hop question answering. In Proceedings of the 2018 conference on empirical methods in natural language processing, p. 2369–2380. Cited by: §6.1. Yao et al. (2023) S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. Narasimhan, and Y. Cao ReAct: synergizing reasoning and acting in language models. In International Conference on Learning Representations, Cited by: §5, §6.1.