Polaris: A G\"odel Agent Framework for Small Language Models through Experience-Abstracted Policy Repair

TL;DR

Polaris is a G"odel agent framework that improves small language models through experience-based policy repair, enabling cumulative learning and consistent performance gains across various benchmarks.

Contribution

It introduces experience abstraction and structured policy repair cycles for small models, enabling self-improvement with auditable, reusable policy patches.

Findings

Polaris improves performance on MGSM, DROP, GPQA, and LitBench benchmarks.

The framework achieves consistent gains over base policies and competitive baselines.

Polaris enables small models to self-correct and refine policies through structured analysis and patching.

Abstract

G\"odel agent realize recursive self-improvement: an agent inspects its own policy and traces and then modifies that policy in a tested loop. We introduce Polaris, a G\"odel agent for compact models that performs policy repair via experience abstraction, turning failures into policy updates through a structured cycle of analysis, strategy formation, abstraction, and minimal code pat ch repair with conservative checks. Unlike response level self correction or parameter tuning, Polaris makes policy level changes with small, auditable patches that persist in the policy and are reused on unseen instances within each benchmark. As part of the loop, the agent engages in meta reasoning: it explains its errors, proposes concrete revisions to its own policy, and then updates the policy. To enable cumulative policy refinement, we introduce experience abstraction, which distills failures into compact, reusable strategies that transfer to unseen instances. On MGSM, DROP, GPQA, and LitBench (covering arithmetic reasoning, compositional inference, graduate-level problem solving, and creative writing evaluation), a 7-billion-parameter model equipped with Polaris achieves consistent gains over the base policy and competitive baselines.