Learning to Repair: Repairing model output errors after deployment using a dynamic memory of feedback

TL;DR

This paper introduces a method for improving large language models post-deployment by using a dynamic memory of user feedback and a corrector model to repair output errors without retraining.

Contribution

It presents FBNet, a system that leverages user feedback stored in memory and a trained corrector to repair model outputs and reduce errors in real-time.

Findings

Up to 30 points improvement in error correction.

Up to 7 points reduction in similar past mistakes.

Demonstrates effectiveness on script generation tasks.

Abstract

Large language models (LMs), while powerful, are not immune to mistakes, but can be difficult to retrain. Our goal is for an LM to continue to improve after deployment, without retraining, using feedback from the user. Our approach pairs an LM with (i) a growing memory of cases where the user identified an output error and provided general feedback on how to correct it (ii) a corrector model, trained to translate this general feedback into specific edits to repair the model output. Given a new, unseen input, our model can then use feedback from similar, past cases to repair output errors that may occur. We instantiate our approach using an existing, fixed model for script generation, that takes a goal (e.g., "bake a cake") and generates a partially ordered sequence of actions to achieve that goal, sometimes containing errors. Our memory-enhanced system, FBNet, learns to apply user feedback to repair such errors (up to 30 points improvement), while making a start at avoiding similar past mistakes on new, unseen examples (up to 7 points improvement in a controlled setting). This is a first step towards strengthening deployed models, potentially broadening their utility. Our code and data is available at https://github.com/allenai/interscript/.