Code Understanding is a Bottleneck for Coding Agents
cs.SE, cs.AI, cs.PL
Submitted: 2026-10-07
Updated: 2026-10-07
Code: https://github.com/microsoft/CABRA
Sources
- gpt-oss-120b & gpt-oss-20b Model Card
- (Im)Paired Programming: Coding Agents Improve Productivity but Harm Understanding
- Evaluating Large Language Models Trained on Code
- Towards Ecologically Valid Research on Language User Interfaces
- LLMs Corrupt Your Documents When You Delegate
- SWE-EVO: Benchmarking Coding Agents in Long-Horizon Software Evolution Scenarios
- Bug In the Code Stack: Can LLMs Find Bugs in Large Python Code Stacks
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- Beyond Resolution Rates: Behavioral Drivers of Coding Agent Success and Failure
- LongCoT: Benchmarking Long-Horizon Chain-of-Thought Reasoning
- How AI Impacts Skill Formation
- OpenAI GPT-5 System Card
- Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks
- Qwen3 Technical Report
- SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering
- TraceLab: Characterizing Coding Agent Workloads for LLM Serving
- Training Versatile Coding Agents in Synthetic Environments
Related papers
- Falsification-Based Verification of LLM-Generated Optimization Models: Sound Test Batteries and Their Detection Limits
- GitSkills: A Dataset of Agent Skills on GitHub
- SABER: Benchmarking Operational Safety of LLM Coding Agents in Stateful Project Workspaces
- PackMonitor: Enabling Zero Package Hallucinations Through Decoding-Time Monitoring
- IntentCoding: Amplifying User Intent in Code Generation
- Incentives and Outcomes in Bug Bounties