Tag
The paper audits frozen LLMs to examine how geometric constraint information is encoded and whether it can be used for generation, influence, and steering, revealing gaps between decodability and actionable outputs.
A blog post exploring how to infer hidden details about frontier AI models' pre-training data and timelines by probing them with curated questions, including estimates of model parameters and training checkpoints.
This paper proposes a framework for probing and steering latent cultural values in LLMs using scenario-based behavioral dilemmas and activation steering, applied across three models and four cultures, finding steerability variation and latent entanglement between cultural dimensions.