Large Multi-modal Models Can Interpret Features in Large Multi-modal Models
This work addresses interpretability for researchers and practitioners using LMMs, but it is incremental as it builds on existing methods like SAEs and applies them to LMMs.
The paper tackles the problem of understanding internal neural representations in Large Multimodal Models (LMMs) by developing a framework that uses Sparse Autoencoders and LMMs themselves to interpret features, demonstrating effective steering of model behavior in tasks like EQ tests.
Recent advances in Large Multimodal Models (LMMs) lead to significant breakthroughs in both academia and industry. One question that arises is how we, as humans, can understand their internal neural representations. This paper takes an initial step towards addressing this question by presenting a versatile framework to identify and interpret the semantics within LMMs. Specifically, 1) we first apply a Sparse Autoencoder(SAE) to disentangle the representations into human understandable features. 2) We then present an automatic interpretation framework to interpreted the open-semantic features learned in SAE by the LMMs themselves. We employ this framework to analyze the LLaVA-NeXT-8B model using the LLaVA-OV-72B model, demonstrating that these features can effectively steer the model's behavior. Our results contribute to a deeper understanding of why LMMs excel in specific tasks, including EQ tests, and illuminate the nature of their mistakes along with potential strategies for their rectification. These findings offer new insights into the internal mechanisms of LMMs and suggest parallels with the cognitive processes of the human brain.