A Mechanistic Interpretation of Arithmetic Reasoning in Language Models using Causal Mediation Analysis
This provides mechanistic insights into arithmetic reasoning in LMs, which is incremental as it builds on existing analysis methods without introducing new paradigms.
The authors tackled the problem of understanding how Transformer-based language models process arithmetic tasks by using causal mediation analysis to identify specific parameters and mechanisms involved, finding that information flows from early layers to the final token via attention and is processed by MLP modules to generate results.
Mathematical reasoning in large language models (LMs) has garnered significant attention in recent work, but there is a limited understanding of how these models process and store information related to arithmetic tasks within their architecture. In order to improve our understanding of this aspect of language models, we present a mechanistic interpretation of Transformer-based LMs on arithmetic questions using a causal mediation analysis framework. By intervening on the activations of specific model components and measuring the resulting changes in predicted probabilities, we identify the subset of parameters responsible for specific predictions. This provides insights into how information related to arithmetic is processed by LMs. Our experimental results indicate that LMs process the input by transmitting the information relevant to the query from mid-sequence early layers to the final token using the attention mechanism. Then, this information is processed by a set of MLP modules, which generate result-related information that is incorporated into the residual stream. To assess the specificity of the observed activation dynamics, we compare the effects of different model components on arithmetic queries with other tasks, including number retrieval from prompts and factual knowledge questions.