Skip to content
#

neurips-paper

Star

Here is 1 public repository matching this topic...

Davide011 / Thesis_public

Star 2

Mechanistic analysis of a GPT-2–like model exploring the compositionality gap in transformers. Using Logit Lens and Causal Tracing, the study identifies and overcomes a deep-layer bottleneck via dataset enhancement addressing the stated Compositionality Gap (NeurIPS24).

tokenizer hpc-clusters explainable-ai ai-research explainability compositionality neurips llm logitlens mechanistic-interpretability llm-training causal-tracing gpt-2-model ai-deployment neurips-paper
  • Updated Oct 29, 2025
  • Python

Improve this page

Add a description, image, and links to the neurips-paper topic page so that developers can more easily learn about it.

Curate this topic

Add this topic to your repo

To associate your repository with the neurips-paper topic, visit your repo's landing page and select "manage topics."

Learn more