Distilling Feedback into Memory-as-a-Tool

Code

• github.com/langchain-ai/deepagents
  OfficialIn paper★ 17,646
• github.com/vicgalle/feedback-memory-as-a-tool
  Official★ 2

Abstract

We propose a framework that amortizes the cost of inference-time reasoning by converting transient critiques into retrievable guidelines, through a file-based memory system and agent-controlled tool calls. We evaluate this method on the Rubric Feedback Bench, a novel dataset for rubric-based learning. Experiments demonstrate that our augmented LLMs rapidly match the performance of test-time refinement pipelines while drastically reducing inference cost.

Reproductions