CoHyDE: Iterative Co-Training of LLM Rewriter & Dense Encoder for Tool Retrieval

• URL: http://arxiv.org/abs/2605.29271v1
• Date: Thu, 28 May 2026 02:41:30 GMT
• Title: CoHyDE: Iterative Co-Training of LLM Rewriter & Dense Encoder for Tool Retrieval
• Abstract summary: CoHyDE trains the dense encoder and the rewriter as a single co-evolving system.<n>On a 10k tool subset of the ToolBench catalog, three rounds of CoHyDE improve over the strongest single-component baseline by +2.5 pp NDCG@5 on standard queries.<n>Using either component in isolation fails to match CoHyDE on both well-formed and vague queries, with losses of up to -8 pp on vague queries.
• Score: 0.3823356975862005
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: Tool retrieval over large API catalogs is a core bottleneck for LLM agents: user queries arrive in colloquial, often underspecified language, while the catalog uses technical API vocabulary that no fixed encoder can bridge on its own. The two dominant training approaches, contrastive encoder fine-tuning and HyDE-style query expansion with a frozen LLM, address this problem from opposite ends and fail in complementary directions: the fine-tuned encoder excels when the query's surface form already matches the catalog but collapses when it does not, while zero-shot HyDE is more robust to underspecified queries yet generates catalog-unaware hypothetical descriptions that degrade retrieval when queries are well-formed. We introduce CoHyDE, an iterative procedure that trains the dense encoder and the LLM rewriter as a single co-evolving system: the encoder is retrained with InfoNCE on catalog-style hypothetical descriptions produced by the rewriter, and the rewriter is preference-aligned via DPO against the encoder's retrieval scores, with both sides warm-started on the tool catalog before the loop begins. On a ~10k tool subset of the ToolBench catalog, three rounds of CoHyDE improve over the strongest single-component baseline by +2.5 pp NDCG@5 on standard queries and +6.3 pp on held-out vague queries, with gains as large as +8 pp on the hardest vague tier. Ablations confirm that co-training is the key ingredient: using either component in isolation fails to match CoHyDE on both well-formed and vague queries, with losses of up to -8 pp on vague queries.

Related papers

• Glyph: A Multi-Strategy Agentic System for Column Description and Sensitivity-Ontology Tagging of Enterprise Data Catalogs [0.0]
  We present Glyph, a production system that frames two coupled problems, column description generation and column type annotation for data classification.<n>The Descriptor grounds generation in the pipeline source code that produces each column, retrieved on demand from an enterprise GitHub.<n>We report end-to-end multi-label tagging quality under a recall-weighted F2 objective across three evaluation groups.
  arXiv  Detail & Related papers  (2026-09-09T16:43:26Z)
• Codebook Agent: Amortized Topology Design for LLM Multi-Agent Systems [67.04448659688579]
  We develop a query-independent 16-entry codebook that ranks the top decoded candidates in a single batched forward pass.<n>With no iterative search and no message passing at test time, Codebook Agent is the most accurate method on all six benchmarks.
  arXiv  Detail & Related papers  (2026-09-02T08:10:22Z)
• It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning [3.106071998555287]
  CoGR trains generators to construct retrieval representations on both query and item sides.<n>Each generator produces a compact set of keywords, which are matched directly through an inverted index.<n>CoGR achieves the best performance on both an internal APP Marketplace dataset and the public WANDS benchmark.
  arXiv  Detail & Related papers  (2026-09-01T03:17:02Z)
• GrepSeek: Training Search Agents for Direct Corpus Interaction [66.69568141699311]
  GrepSeek is an optimized direct corpus interaction (DCI) search agent that trains a compact search agent to find, filter, and compose evidence from large text corpora.<n>To make DCI practical at scale, we further use a semantics-preserving sharded-parallel execution engine that accelerates shell-based retrieval by up to $7.6times$ while preserving byte-exact equivalence with sequential execution of the shell command.
  arXiv  Detail & Related papers  (2026-05-28T03:37:33Z)
• Retrieval as Reasoning: Self-Evolving Agent-Native Retrieval via LLM-Wiki [6.956097396264084]
  LLM agents require retrieval to behave less like one-shot context fetching and more like reasoning.<n>Current Retrieval-Augmented Generation (RAG) systems organize external knowledge as flat chunks retrieved by embedding similarity.<n>We propose LLM-Wiki, an agent-native retrieval system that operationalizes the Retrieval-as-Reasoning paradigm by treating external knowledge as a compilable, composable, and self-evolving structure.
  arXiv  Detail & Related papers  (2026-05-25T06:36:14Z)
• Domain-Adaptive and Scalable Dense Retrieval for Content-Based Recommendation [0.0]
  We present a scalable dense retrieval system based on a two-tower bi-encoder, fine-tuned on the Amazon Reviews 2023 (Fashion) subset.<n>We construct training pairs from review text (as a query proxy) and item metadata (as the positive document) and fine-tune on 50,000 sampled interactions with a maximum sequence length of 500 tokens.<n>On a review-to-title benchmark over 826,402 catalog items, our approach improves Recall@10 from 0.26 (BM25) to 0.66.
  arXiv  Detail & Related papers  (2026-01-31T20:58:23Z)
• Succeeding at Scale: Automated Multi-Retriever Fusion and Query-Side Adaptation for Multi-Tenant Search [22.080200394842123]
  DevRev Search is a passage retrieval benchmark for technical customer support constructed through a fully automatic pipeline.<n>We propose a practical textbfIndex-Preserving Adaptation strategy, by fine-tuning only the query encoder via Low-Rank Adaptation.<n>Our experiments on DevRev Search and SciFact demonstrate that targeting specific transformer layers in the query encoder yields optimal quality-efficiency trade-offs.
  arXiv  Detail & Related papers  (2026-01-08T06:44:40Z)
• SweRank: Software Issue Localization with Code Ranking [109.3289316191729]
  SweRank is an efficient retrieve-and-rerank framework for software issue localization.<n>We construct SweLoc, a large-scale dataset curated from public GitHub repositories.<n>We show that SweRank achieves state-of-the-art performance, outperforming both prior ranking models and costly agent-based systems.
  arXiv  Detail & Related papers  (2025-05-07T19:44:09Z)
• Effective Instruction Parsing Plugin for Complex Logical Query Answering on Knowledge Graphs [51.33342412699939]
  Knowledge Graph Query Embedding (KGQE) aims to embed First-Order Logic (FOL) queries in a low-dimensional KG space for complex reasoning over incomplete KGs. Recent studies integrate various external information (such as entity types and relation context) to better capture the logical semantics of FOL queries. We propose an effective Query Instruction Parsing (QIPP) that captures latent query patterns from code-like query instructions.
  arXiv  Detail & Related papers  (2024-10-27T03:18:52Z)
• RRADistill: Distilling LLMs' Passage Ranking Ability for Long-Tail Queries Document Re-Ranking on a Search Engine [2.0379810233726126]
  Large Language Models (LLMs) excel at understanding the semantic relationships between queries and documents. These queries are challenging for feedback-based rankings due to sparse user engagement and limited feedback. We propose an efficient label generation pipeline and novel sLLM training methods for both encoder and decoder models.
  arXiv  Detail & Related papers  (2024-10-08T11:28:06Z)
• List-aware Reranking-Truncation Joint Model for Search and Retrieval-augmented Generation [80.12531449946655]
  We propose a Reranking-Truncation joint model (GenRT) that can perform the two tasks concurrently. GenRT integrates reranking and truncation via generative paradigm based on encoder-decoder architecture. Our method achieves SOTA performance on both reranking and truncation tasks for web search and retrieval-augmented LLMs.
  arXiv  Detail & Related papers  (2024-02-05T06:52:53Z)
• Noise-Robust Dense Retrieval via Contrastive Alignment Post Training [89.29256833403167]
  Contrastive Alignment POst Training (CAPOT) is a highly efficient finetuning method that improves model robustness without requiring index regeneration. CAPOT enables robust retrieval by freezing the document encoder while the query encoder learns to align noisy queries with their unaltered root. We evaluate CAPOT noisy variants of MSMARCO, Natural Questions, and Trivia QA passage retrieval, finding CAPOT has a similar impact as data augmentation with none of its overhead.
  arXiv  Detail & Related papers  (2023-04-06T22:16:53Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.