Causal Abstractions of Neural Networks

Authors: Atticus Geiger, Hanson Lu, Thomas Icard, Christopher Potts

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We apply this method in a case study to analyze neural models trained on Multiply Quantified Natural Language Inference (MQNLI) corpus, a highly complex NLI dataset that was constructed with a tree-structured natural logic causal model.
Researcher Affiliation Academia Atticus Geiger , Hanson Lu , Thomas Icard, and Christopher Potts Stanford Stanford, CA 94305-2150 {atticusg, hansonlu, icard, cgpotts}@stanford.edu
Pseudocode No The paper does not contain any pseudocode or algorithm blocks.
Open Source Code Yes We provide tools for causal abstraction analysis at http://github.com/hansonhl/antra and the code base for this paper at http://github.com/atticusg/Interchange
Open Datasets Yes The Multiply Quantified NLI (MQNLI) dataset of Geiger et al. [10] contains templatically generated English-language NLI examples...
Dataset Splits Yes MQNLI has train/dev/test splits that vary in their difficulty.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running experiments.
Software Dependencies No The paper mentions BERT and Huggingface's transformers but does not provide specific version numbers for these or other software dependencies.
Experiment Setup No Additional model and training details are given in Appendix B. (This indicates details exist but are not in the provided main text).