mlsanityData to 14 Oct 2026

DeepDiscovery: A Location-Inference Framework for Task-Level Repository Understanding

arXiv 2606.22906 ↗First posted 22 Jun 2026cs.SE cs.AIcited by 1

Not on the map. Why?

Abstract (arXiv)

Large language models have shown strong performance on software engineering (SE) tasks, yet understanding large industrial repositories remains challenging. Existing methods often retrieve only local fragments and fail to recover the broader task-relevant context needed for complex repository-level tasks. We present DeepDiscovery, a task-level repository-understanding method for large industrial codebases. DeepDiscovery uses a two-stage textitLocation--Inference framework to localize high-confidence task anchors and recover broader task-relevant context over multi-relational repository structure under budget constraints. Across controlled method-level evaluation, organization-internal industrial repository-understanding scenarios, and end-to-end evaluation on SWE-bench Verified, DeepDiscovery consistently improves task-relevant file recovery and downstream SE performance. On 27 medium-scale tasks, DeepDiscovery achieves the best file recovery quality among five representative baselines without offline preprocessing. On organization-internal industrial tasks from a production-scale integrated codebase ecosystem, including 27 medium-scale tasks and 40 large-scale tasks, DeepDiscovery improves Full Recall Rate across multiple AI coding systems, with absolute gains ranging from 1.6 to 9.2 percentage points on large subprojects and from 2.5 to 7.4 percentage points on medium-scale subprojects. In a controlled end-to-end evaluation on SWE-bench Verified, a system equipped with DeepDiscovery achieves a 78.6% Solve Rate, outperforming the corresponding baseline by 8.2 percentage points. These results suggest that stronger task-level repository understanding can improve coding-agent performance on complex SE tasks.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 31 of its 38 references that are in the map. First posted on arXiv 22 Jun 2026.

Builds on

31 of 38 references.

References · 31

  1. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,373
  2. 2024From Local to Global: A Graph RAG Approach to Query-Focused Summarizationcited by 1,453
  3. 2026Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfacescited by 585
  4. 2024CodeAgent: Enhancing Code Generation with Tool-Integrated Agent Systems for Real-World Repo-level Coding Challengescited by 282
  5. 2024Advancing LLM Reasoning Generalists with Preference Treescited by 192
  6. 2024EASYTOOL: Enhancing LLM-based Agents with Concise Tool Instructioncited by 154
  7. 2024RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrievalcited by 116
  8. 2024A Review on Edge Large Language Models: Design, Execution, and Applicationscited by 100
  9. 2024BurstGPT: A Real-world Workload Dataset to Optimize LLM Serving Systemscited by 87
  10. 2024Vul-RAG: Enhancing LLM-based Vulnerability Detection via Knowledge-level RAGcited by 68
  11. 2024Alibaba LingmaAgent: Improving Automated Issue Resolution via Comprehensive Repository Explorationcited by 58
  12. 2025A Survey of Large Language Model Agents for Question Answeringcited by 48
19 more
  1. 2023Make LLM a Testing Expert: Bringing Human-like Interaction to Mobile GUI Testing via Functionality-aware Decisionscited by 41
  2. 2025GFM-RAG: Graph Foundation Model for Retrieval Augmented Generationcited by 40
  3. 2024A Systematic Literature Review on Large Language Models for Automated Program Repaircited by 37
  4. 2024Source Code Summarization in the Era of Large Language Modelscited by 31
  5. 2025RepoMaster: Autonomous Exploration and Understanding of GitHub Repositories for Complex Task Solvingcited by 19
  6. 2025An Empirical Study of Retrieval-Augmented Code Generation: Challenges and Opportunitiescited by 18
  7. 2025DependEval: Benchmarking LLMs for Repository Dependency Understandingcited by 17
  8. 2025GraphCodeAgent: Dual Graph-Guided LLM Agent for Retrieval-Augmented Repo-Level Code Generationcited by 17
  9. 2025LoCoBench-Agent: An Interactive Benchmark for LLM Agents in Long-Context Software Engineeringcited by 16
  10. 2024Retrieve-Plan-Generation: An Iterative Planning and Answering Framework for Knowledge-Intensive LLM Generationcited by 15
  11. 2025Empowering GraphRAG with Knowledge Filtering and Integrationcited by 15
  12. 2025AgentInit: Initializing LLM-based Multi-Agent Systems via Diversity and Expertise Orchestration for Effective and Efficient Collaborationcited by 10
  13. 2025LooGLE v2: Are LLMs Ready for Real World Long Dependency Challenges?cited by 8
  14. 2026ProcCtrlBench: Evaluating Process-Level Defects and Control Preservation in LLM Coding Agentscited by 6
  15. 2026Reliable Graph-RAG for Codebases: AST-Derived Graphs vs LLM-Extracted Knowledge Graphscited by 4
  16. 2026Beyond Code Snippets: Benchmarking LLMs on Repository-Level Question Answeringcited by 3
  17. 2026ArchAgent: Scalable Legacy Software Architecture Recovery with LLMscited by 2
  18. 2026LogicScan: An LLM-driven Framework for Detecting Business Logic Vulnerabilities in Smart Contractscited by 2
  19. 2025Vector Graph-Based Repository Understanding for Issue-Driven File Retrievalcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026The Recall Trap: A Recall-Maximizing Retriever Configuration Reduces Issue Resolution in Fixed-Budget Code Contextcited by 0