Fine-tuning language models~(LMs) on human-generated data remains a prevalent
practice. However, the performance of such models is often limited by the quantity and …

Autoregressive large language models (LLMs) compress knowledge from their training data
through next-token conditional distributions. This limits tractable querying of this knowledge …

When prompting language models to complete a task, users often leave important aspects
unsaid. While asking questions could resolve this ambiguity\citep [GATE;][]{li2023eliciting} …

When writing and talking, people sometimes pause to think. Although reasoning-focused
works have often framed reasoning as a method of answering questions or completing …

We build a computational model of how humans actively infer hidden rules by doing
experiments. The basic principles behind the model is that, even if the rule is deterministic …

A fundamental skill among human developers is the ability to understand and reason about
program execution. As an example, a programmer can mentally simulate code execution in …

Is there a way to design powerful AI systems based on machine learning methods that would
satisfy probabilistic safety guarantees? With the long-term goal of obtaining a probabilistic …

Chain-of-Thought (CoT) reasoning could in principle enable a deeper understanding of a
language model's (LM) internal reasoning. However, prior work suggests that some LMs …

Faithful and interpretable reasoning in language models can be achieved by imposing a
bottleneck on the model: generating explanatory notes on how to solve a task, and then …