Sufficient Context: A New Lens on Retrieval Augmented Generation Systems

2024-11-09

arXiv.org

AI Summary

This work investigates the root causes of LLM errors in RAG systems: whether failures stem from underutilization of retrieved context or from intrinsic insufficiency of the context itself. To this end, we formally define and quantify “sufficient context” for a given query, and build a binary classifier to determine whether retrieved passages contain adequate information to answer the query. We conduct cross-model error stratification, revealing that large models tend to generate incorrect answers rather than abstain, whereas smaller models exhibit higher rates of hallucination or excessive abstention. Leveraging these insights, we propose a context-sufficiency-aware controllable generation framework featuring a guided abstention mechanism. Experiments across Gemini, GPT, and Gemma show 2–10% absolute accuracy gains and substantial hallucination reduction. Our code, evaluation prompts, and key findings are publicly released.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Safety and Robustness
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Augmenting LLMs with context leads to improved performance across many applications. Despite much research on Retrieval Augmented Generation (RAG) systems, an open question is whether errors arise because LLMs fail to utilize the context from retrieval or the context itself is insufficient to answer the query. To shed light on this, we develop a new notion of sufficient context, along with a method to classify instances that have enough information to answer the query. We then use sufficient context to analyze several models and datasets. By stratifying errors based on context sufficiency, we find that larger models with higher baseline performance (Gemini 1.5 Pro, GPT 4o, Claude 3.5) excel at answering queries when the context is sufficient, but often output incorrect answers instead of abstaining when the context is not. On the other hand, smaller models with lower baseline performance (Mistral 3, Gemma 2) hallucinate or abstain often, even with sufficient context. We further categorize cases when the context is useful, and improves accuracy, even though it does not fully answer the query and the model errs without the context. Building on our findings, we explore ways to reduce hallucinations in RAG systems, including a new selective generation method that leverages sufficient context information for guided abstention. Our method improves the fraction of correct answers among times where the model responds by 2--10% for Gemini, GPT, and Gemma. Key findings and the prompts used in our autorater analysis are available on our github.

Problem

• Determining if RAG errors stem from insufficient context or poor LLM utilization
• Analyzing model performance based on context sufficiency across different datasets
• Reducing hallucinations in RAG systems via selective generation methods

Innovation

• Introduces sufficient context classification method
• Analyzes models using context sufficiency stratification
• Develops selective generation for guided abstention