SGAgent: Suggestion-Guided LLM-Based Multi-Agent Framework for Repository-Level Software Repair
2026-03-04T20:21:24Z•56a24d72a09e7cdbf9f32e0afd2bdb68689dc5672a713d796f9480cc0868bfa9
IC3/PDRKV-cache-memoryLLM-based-repairSWE-rebenchdatasetsembodied-planningfailure-attributionflaky-testsformal-verificationfunctional-test-generationfuzzingmicroservice-testingmulti-agent-systemsreinforcement-learningsecure-code-generationseed-selectionvulnerability-repair
What happened
This collection of 2026 software-engineering and systems preprints highlights advances in LLM-driven developer tooling, testing, and verification. Key contributions: SGAgent — a suggestion-guided multi-agent localize-suggest-fix framework that improves repository- and vulnerability-level repair (notably 48% accuracy on VUL4J/VJBench); Vul2Safe/SRCode — methods and a PrimeVul+ dataset for constructing repair pairs and using token-level RL rewards to reduce insecure code generation; PoCo — a technique for improving coverage-based fuzzing seed selection; SWE-rebench V2 — a large language-agnostic
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_cs_se
- Record identifier
- 56a24d72a09e7cdbf9f32e0afd2bdb68689dc5672a713d796f9480cc0868bfa9
- Enrichment time
- 2026-03-04T20:21:24Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.