7 accepted papers
Existing approaches for manipulating neural network activations, such as PCA and SAEs, rely on strong assumptions about activation structure. We develop a generative approach that models activations with diffusion, that makes minimal assumptions and improves with data and model scale. We use this ac…
Pretrained language models (LMs) can generalize to implications of facts that they are finetuned on. For example, if finetuned on "John Doe lives in Tokyo," LMs correctly answer "What language do the people in John Doe's city speak?'' with "Japanese''. However, little is known about the mechanisms t…
Recent methods in language model interpretability employ techniques such as sparse autoencoders to decompose residual stream contributions into linear, semantically-meaningful features. Our work demonstrates that an underlying assumption of these methods—that residual stream contributions build addi…
Language models (LMs) are susceptible to bias, sycophancy, backdoors, and other tendencies that lead to unfaithful responses to the input context. Interpreting internal states of LMs could help monitor and correct unfaithful behavior. We hypothesize that LMs faithfully represent their input contexts…
Language models (LMs) can recall facts mentioned in context, as shown by their performance on reading comprehension tasks. When the context describes facts about more than one entity, the LM has to correctly bind attributes to their corresponding entity. We show, via causal experiments, that LMs' in…
Effective planning in the real world requires not only world knowledge, but the ability to leverage that knowledge to build the right representation of the task at hand. Decades of hierarchical planning techniques have used domain-specific temporal action abstractions to support efficient and accura…
We present an improved method for symbolic regression that seeks to fit data to formulas that are Pareto-optimal, in the sense of having the best accuracy for a given complexity. It improves on the previous state-of-the-art by typically being orders of magnitude more robust toward noise and bad data…