What Coding Agents Leave Unsaid: Measuring Task-State Gaps Between Execution and Reports
cs.AI
Submitted: 2026-10-14
Updated: 2026-10-15
Code: https://github.com/Kaguraaaaaa/Ctxgap
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sources
- ContextBench: A Benchmark for Context Retrieval in Coding Agents
- Code as Agent Harness
- AgentLens: Production-Assessed Trajectory Reviews for Coding Agent Evaluation
- KnowledgeGain: Evaluating and Optimizing Science News Generation for Reader Learning
- How Coding Agents Fail Their Users: A Large-Scale Analysis of Developer-Agent Misalignment in 20,574 Real-World Sessions
- OpenHands: An Open Platform for AI Software Developers as Generalist Agents
- The Observability Gap: Why Output-Level Human Feedback Fails for LLM Coding Agents
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems