Kintsugi: Learning Policies by Repairing Executable Knowledge Bases

TL;DR

Kintsugi is a white-box framework for improving embodied policies by constructing and editing executable, type-checked knowledge bases, enabling better inspection, validation, and reuse without relying on language models during inference.

Contribution

It introduces a novel knowledge base representation and verifier-gated editing process for policy learning, enhancing interpretability and local editability in embodied agents.

Findings

Achieves strong performance on long-horizon benchmarks.

Maintains inspectability and local editability of policies.

Operates without language model calls during inference.

Abstract

Modern embodied agents achieve impressive performance, but their task knowledge is often stored in neural weights, latent state, or prompt-bound memory, making individual policy knowledge difficult to inspect, validate, recombine, and reuse. We introduce \textbf{Kintsugi}, a white-box policy-learning framework that treats embodied policy improvement as verifier-gated construction of a typed executable Knowledge Base (KB). Kintsugi represents task-level policy knowledge as composable typed entries -- predicates, operators, policy schemas, monitors, recovery rules, experience records, and goals -- and improves this artifact through localized typed edits induced from rollout evidence, rather than relying on test-time language-model reasoning. Between rollouts, a tool-constrained agentic editing loop diagnoses trajectory failures, localizes them to editable KB layers, and proposes candidate edits. A deterministic verification gate admits an edit only when the candidate type-checks, the resulting KB executes, and focused validation success or trajectory-health metrics improve without violating protected-regression checks. At inference, the accepted KB is executed by a deterministic symbolic executor with zero LLM calls. Across long-horizon text-agent benchmarks and representative object-centric manipulation settings, Kintsugi achieves strong endpoint performance while preserving inspectability, local editability, and verifier-gated deployment. These results suggest that embodied policy improvement can be organized around executable task knowledge.