Showing cs.LGShow all
2 papers · 1 filter
cs.LG2024
Extracting Interpretable Task-Specific Circuits from Large Language Models for Faster Inference
Jorge GarcÃa-Carrasco, Alejandro Maté, Juan Trujillo
Large Language Models (LLMs) have shown impressive performance across a wide range of tasks. However, the size of LLMs is steadily increasing, hindering their application on comput…
cs.LG2024
How does GPT-2 Predict Acronyms? Extracting and Understanding a Circuit via Mechanistic Interpretability
Jorge GarcÃa-Carrasco, Alejandro Maté, Juan Trujillo
Transformer-based language models are treated as black-boxes because of their large number of parameters and complex internal interactions, which is a serious safety concern. Mecha…