Activity (Process)

Bounded Diagnostic Code Repair in JIT-Agent

In JIT-Agent, bounded diagnostic code repair is an error recovery mechanism where the generator learns to ingest diagnostic execution feedback, synthesize structured local patches from a patch space P\mathcal{P}, and apply them deterministically to an unstable harness. Rather than attempting open-ended or unconstrained debugging, the repair process is restricted to a maximum of two revision rounds (K⋆≤2K^\star \le 2). This short-horizon constraint ensures the model focuses on high-leverage, locally recoverable failures—such as compiler errors, interface mismatches, tool-call failures, and runtime exceptions—preventing compounding errors, high latency, and excessive token expenditure during test-time execution.

0

1

Updated 2026-10-02

Tags

Prep Sessions

Frontier Foundation Models, Capability Evaluation, and Just-In-Time Agent Harnesses @ University of Michigan - Ann Arbor

Ch.3 Adaptive Agent Harness Design - Frontier Foundation Models, Capability Evaluation, and Just-In-Time Agent Harnesses @ University of Michigan - Ann Arbor

Stage II Error Recovery via Bounded Diagnostic Repair - Frontier Foundation Models, Capability Evaluation, and Just-In-Time Agent Harnesses @ University of Michigan - Ann Arbor

Dynamic Agent Scaffolding: Synthesis, Diagnostic Repair, and Evolutionary Optimization @ University of Michigan - Ann Arbor

Ch.2 Multi-Stage Harness Optimization - Dynamic Agent Scaffolding: Synthesis, Diagnostic Repair, and Evolutionary Optimization @ University of Michigan - Ann Arbor

Stage II: Diagnostic Feedback and Bounded Harness Repair - Dynamic Agent Scaffolding: Synthesis, Diagnostic Repair, and Evolutionary Optimization @ University of Michigan - Ann Arbor