7 papers
Compact Path Representation in DAGs via Colored Edge Pebbling
Paola Bonizzoni, Alessio Conte, Gianluca Della Vedova +3
Compactly representing a variation graph is a core problem in computational pangenomics that is usually attacked with techniques that have been originated on texts and adapted to g…
Narrow Finetuning Leaves Clearly Readable Traces in Activation Differences
Julian Minder, Clément Dumas, Stewart Slocum +4
Finetuning on narrow domains has become an essential tool to adapt Large Language Models (LLMs) to specific tasks and to create models with known unusual properties that are useful…
Overcoming Sparsity Artifacts in Crosscoders to Interpret Chat-Tuning
Julian Minder, Clément Dumas, Caden Juang +2
Model diffing is the study of how fine-tuning changes a model's representations and internal algorithms. Many behaviors of interest are introduced during fine-tuning, and model dif…
Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers
Adam Karvonen, James Chua, Clément Dumas +8
Large language model (LLM) activations are notoriously difficult to understand, with most existing techniques using complex, specialized methods for interpreting them. Recent work…
The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability?
Denis Sutter, Julian Minder, Thomas Hofmann +1
The concept of causal abstraction got recently popularised to demystify the opaque decision-making processes of machine learning models; in short, a neural network can be abstracte…
Believe It or Not: How Deeply do LLMs Believe Implanted Facts?
Stewart Slocum, Julian Minder, Clément Dumas +4
Knowledge editing techniques promise to implant new factual knowledge into large language models (LLMs). But do LLMs really believe these facts? We develop a framework to measure b…