GRADE: Probing Knowledge Gaps in LLMs through Gradient Subspace Dynamics
This addresses the challenge of responsible LLM deployment by providing a more reliable way to identify knowledge gaps, though it is incremental as it builds on existing internal probing methods.
The paper tackles the problem of detecting when a large language model lacks sufficient internal knowledge to answer a question correctly, proposing GRADE, a method that uses gradient subspace dynamics to quantify knowledge gaps, and validates it on six benchmarks with demonstrated effectiveness and robustness.
Detecting whether a model's internal knowledge is sufficient to correctly answer a given question is a fundamental challenge in deploying responsible LLMs. In addition to verbalising the confidence by LLM self-report, more recent methods explore the model internals, such as the hidden states of the response tokens to capture how much knowledge is activated. We argue that such activated knowledge may not align with what the query requires, e.g., capturing the stylistic and length-related features that are uninformative for answering the query. To fill the gap, we propose GRADE (Gradient Dynamics for knowledge gap detection), which quantifies the knowledge gap via the cross-layer rank ratio of the gradient to that of the corresponding hidden state subspace. This is motivated by the property of gradients as estimators of the required knowledge updates for a given target. We validate \modelname{} on six benchmarks, demonstrating its effectiveness and robustness to input perturbations. In addition, we present a case study showing how the gradient chain can generate interpretable explanations of knowledge gaps for long-form answers.