Field Notes · Closing

When Does Bigger Help? A Controlled Study of LLM Scale for Ontology Learning

This paper evaluates how LLM size affects ontology learning performance across 13 models from different families using a retrieval-augmented generation pipeline.

BLOOM-WILT: Logit Tilting for Behaviour Elicitation in Automated LLM Auditing

This paper proposes a method for improving the efficiency of automated LLM auditing by using logit tilting to increase the likelihood of discovering problematic model behaviors during testing.

S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?

S3Gym is a framework that investigates whether large language models can autonomously test their own behavior, evaluate the results, and use those evaluations to improve their decision-making in agent tasks.