Accepted but Wrong: Silent Failures of LLM Code Repairs after API Migrations
cs.SE, cs.AI
Submitted: 2026-11-05
Updated: 2026-11-13
Terminology
Sources
- CodeT: Code Generation with Generated Tests
- Bench2Robust: Scenario-Controlled Recovery for Tool-Using Agents
- Teaching Large Language Models to Self-Debug
- To Diff or Not to Diff? Structure-Aware and Adaptive Output Formats for Efficient LLM-based Code Editing
- ContractRL: Shielded Group-Relative Policy Optimization for Auditable Tool-Call Repair
- Beyond Task Completion: A Verification-vs.-Conformance Gap in Tool-Evolving Agents
- DynamicMCPBench: A Trace-Grounded, Effect-Scored Benchmark for LLM Agents over Live MCP Servers
- Loud Failures, Quiet Failures: Fault Detection and Recovery in Tool-Using Language Model Agents
- CodeUpdateArena: Benchmarking Knowledge Editing on API Updates
- Client--Library Compatibility Testing with API Interaction Snapshots
- Can Large Language Models Write Good Property-Based Tests?
- PAFT: Preservation Aware Fine-Tuning for Minimal-Edit Program Repair
- DepRepair: LLM-Based Source-Code Repair for Dependency Breaking Changes
- Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
- When Models Edit Too Much: On the Fidelity of Minimal Code Edits
Related papers
- ArkEval: Benchmarking and Evaluating Automated CodeRepair for ArkTS
- TestPrism: Rethinking Test Evaluation Beyond a Single Reference
- Mining Workflow Graphs for Black-Box Boundary Testing of Conversational LLM Agents
- Falsification-Based Verification of LLM-Generated Optimization Models: Sound Test Batteries and Their Detection Limits
- GitSkills: A Dataset of Agent Skills on GitHub
- For Coding Agents, the Repository Matters