A Glitch in the Matrix? Locating and Detecting Language Model Grounding with Fakepedia
This addresses the challenge of improving retrieval-augmented generation methods for AI researchers and practitioners, though it is incremental as it builds on existing knowledge of factual recall mechanisms.
The paper tackled the problem of understanding how large language models (LLMs) ground contextual information, especially when it contradicts their stored knowledge, by introducing Fakepedia, a counterfactual dataset, and using causal mediation analysis to identify computational patterns; they found that distinguishing grounded from ungrounded responses is achievable through computational analysis.
Large language models (LLMs) have an impressive ability to draw on novel information supplied in their context. Yet the mechanisms underlying this contextual grounding remain unknown, especially in situations where contextual information contradicts factual knowledge stored in the parameters, which LLMs also excel at recalling. Favoring the contextual information is critical for retrieval-augmented generation methods, which enrich the context with up-to-date information, hoping that grounding can rectify outdated or noisy stored knowledge. We present a novel method to study grounding abilities using Fakepedia, a novel dataset of counterfactual texts constructed to clash with a model's internal parametric knowledge. In this study, we introduce Fakepedia, a counterfactual dataset designed to evaluate grounding abilities when the internal parametric knowledge clashes with the contextual information. We benchmark various LLMs with Fakepedia and conduct a causal mediation analysis of LLM components when answering Fakepedia queries, based on our Masked Grouped Causal Tracing (MGCT) method. Through this analysis, we identify distinct computational patterns between grounded and ungrounded responses. We finally demonstrate that distinguishing grounded from ungrounded responses is achievable through computational analysis alone. Our results, together with existing findings about factual recall mechanisms, provide a coherent narrative of how grounding and factual recall mechanisms interact within LLMs.