Training Language Models to Explain Their Own Computations
This addresses the interpretability challenge in AI by enabling models to generate scalable explanations of their own behavior, though it is incremental as it builds on existing techniques.
The study tackled the problem of whether language models can learn to faithfully describe their own internal computations, such as feature information and causal structures, and found that fine-tuned explainer models show non-trivial generalization, with self-explanation generally outperforming explanations from different models.
Can language models (LMs) learn to faithfully describe their internal computations? Are they better able to describe themselves than other models? We study the extent to which LMs' privileged access to their own internals can be leveraged to produce new techniques for explaining their behavior. Using existing interpretability techniques as a source of ground truth, we fine-tune LMs to generate natural language descriptions of (1) the information encoded by LM features, (2) the causal structure of LMs' internal activations, and (3) the influence of specific input tokens on LM outputs. When trained with only tens of thousands of example explanations, explainer models exhibit non-trivial generalization to new queries. This generalization appears partly attributable to explainer models' privileged access to their own internals: using a model to explain its own computations generally works better than using a *different* model to explain its computations (even if the other model is significantly more capable). Our results suggest not only that LMs can learn to reliably explain their internal computations, but that such explanations offer a scalable complement to existing interpretability methods.