Beyond Semantic Similarity: Rethinking Retrieval for Agentic Search via Direct Corpus Interaction
| Field | Content | |---|---| | Authors | Zhuofeng Li, Haoxiang Zhang, Cong Wei, Pan Lu, Ping Nie, Yi Lu, et al. (19 authors) | | Published | 2026-05-03 | | Source | https://arxiv.org/abs/2605.05242 | | Type | Research paper — Agentic Search |
TL;DR
Retrieval for agentic search should not be reduced to a fixed top-k similarity interface. The paper studies direct corpus interaction (DCI): an agent searches the raw corpus directly with general-purpose terminal tools (grep, file reads, shell commands, lightweight scripts) — no embedding model, no vector index, no retrieval API — and outperforms strong sparse, dense, and reranking baselines on BRIGHT, BEIR, BrowseComp-Plus, and multi-hop QA tasks.
Background and Motivation
Modern retrieval systems, whether lexical or semantic, expose a corpus through a fixed similarity interface that compresses access into a single top-k retrieval step before reasoning. This abstraction is efficient, but for agentic search it becomes a bottleneck:
- Exact lexical constraints are hard to express via similarity scores.
- Sparse clue conjunctions and local context checks require operations an off-the-shelf retriever does not support.
- Evidence filtered out early cannot be recovered by stronger downstream reasoning.
- Agentic tasks (discovering intermediate entities, combining weak clues, revising plans after partial evidence) require multi-step orchestration that a one-shot retriever cannot serve.
- Tools: grep, file reads, shell commands, lightweight scripts
- No offline indexing: no embedding model, vector index, or retrieval API
- Adaptability: works naturally on evolving local corpora
- Substantially outperforms strong sparse, dense, and reranking baselines on several BRIGHT and BEIR datasets.
- Attains strong accuracy on BrowseComp-Plus and multi-hop QA — without relying on any conventional semantic retriever.
- A Systematic Framework for Enterprise Knowledge Retrieval
- RAG and Understanding in Vision: A Survey
- Synergizing RAG and Reasoning: A Systematic Review
- AceSearcher: Bootstrapping Reasoning and Search for LLMs via RL
- AgentX: Agent-Driven Self-Iteration of Industrial Recommender Systems
- Agentic Information Retrieval (Oct 2024, arXiv)
Method: Direct Corpus Interaction (DCI)
DCI lets the agent interact with the raw corpus directly using general-purpose terminal tools:
This replaces the fixed similarity interface with a higher-resolution interface through which the agent queries and inspects the corpus at arbitrary granularity.
Results
Key Insight
As language agents become stronger, retrieval quality depends not only on reasoning ability but also on the resolution of the interface through which the model interacts with the corpus. DCI opens a broader interface-design space for agentic search.
Original Abstract
> Modern retrieval systems, whether lexical or semantic, expose a corpus through a fixed similarity interface that compresses access into a single top-k retrieval step before reasoning. This abstraction is efficient, but for agentic search, it becomes a bottleneck: exact lexical constraints, sparse clue conjunctions, local context checks, and multi-step hypothesis refinement are difficult to implement by calling a conventional off-the-shelf retriever, and evidence filtered out early cannot be recovered by stronger downstream reasoning. Agentic tasks further exacerbate this limitation because they require agents to orchestrate multiple steps, including discovering intermediate entities, combining weak clues, and revising the plan after observing partial evidence. To tackle the limitation, we study direct corpus interaction (DCI), where an agent searches the raw corpus directly with general-purpose terminal tools (e.g., grep, file reads, shell commands, lightweight scripts), without any embedding model, vector index, or retrieval API. This approach requires no offline indexing and adapts naturally to evolving local corpora. Across IR benchmarks and end-to-end agentic search tasks, this simple setup substantially outperforms strong sparse, dense, and reranking baselines on several BRIGHT and BEIR datasets, and attains strong accuracy on BrowseComp-Plus and multi-hop QA without relying on any conventional semantic retriever. Our results indicate that as language agents become stronger, retrieval quality depends not only on reasoning ability but also on the resolution of the interface through which the model interacts with the corpus, with which DCI opens a broader interface-design space for agentic search.