STACKFEED: Structured Textual Actor-Critic Knowledge Base Editing with FeedBack
This addresses the issue of unreliable information generation by LLMs for users in low-resource domains like programming and factual QA, though it appears incremental as it builds on existing RAG and reinforcement learning methods.
The paper tackles the problem of inaccuracies in knowledge bases used by Retrieval-Augmented Generation systems, especially in low-resource settings, by introducing STACKFEED, which iteratively refines knowledge bases using expert feedback and a reinforcement learning framework, resulting in significant improvements in knowledge base quality and system performance.
Large Language Models (LLMs) often generate incorrect or outdated information, especially in low-resource settings or when dealing with private data. To address this, Retrieval-Augmented Generation (RAG) uses external knowledge bases (KBs), but these can also suffer from inaccuracies. We introduce STACKFEED, a novel Structured Textual Actor-Critic Knowledge base editing with FEEDback approach that iteratively refines the KB based on expert feedback using a multi-actor, centralized critic reinforcement learning framework. STACKFEED defines a ReACT actor agent on each document to perform structured edits based on document specific targeted instructions. Experimental results showcase that STACKFEED significantly improves KB quality and performance of the RAG system. We evaluate STACKFEED on low-resource programming problems, modified python packaged and factual question-answering tasks.