This work identifies 18 foundational challenges in assuring the alignment and safety of large
language models (LLMs). These challenges are organized into three different categories …

Large language models (LLMs) appear to bias their survey answers toward certain values.
Nonetheless, some argue that LLMs are too inconsistent to simulate particular values. Are …

Language models (LMs) have been shown to behave unexpectedly post-deployment. For
example, new jailbreaks continually arise, allowing model misuse, despite extensive red …

The use of large language models (LLMs) as automated evaluation tools to assess the
quality of generated natural language, known as LLMs-as-Judges, has demonstrated …

Humans acquire knowledge by observing the external world, but also by introspection.
Introspection gives a person privileged access to their current state of mind (eg, thoughts …

Models trained specifically to generate long Chains of Thought (CoTs) have recently
achieved impressive results. We refer to these models as Inference-Time-Compute (ITC) …

Language models are known to absorb biases from their training data, leading to predictions
driven by statistical regularities rather than semantic relevance. We investigate the impact of …

Advances in Artificial Intelligence have brought about an evolution of how reasoning has
been developed for modern AI models. I show how the process of human reinforcement has …