Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Quantifying Consistency and Information Loss for Causal Abstraction Learning

Authors: Fabio Massimo Zennaro, Paolo Turrini, Theodoros Damoulas

IJCAI 2023 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we illustrate the flexibility of our setup by empirically showing how different measures and algorithmic choices may lead to different abstractions. We run empirical simulations for the two scenarios in Fig. 2
Researcher Affiliation Academia Fabio Massimo Zennaro , Paolo Turrini and Theodoros Damoulas University of Warwick,Coventry, United Kingdom EMAIL
Pseudocode Yes Algorithm 1 Overall IC error evaluation; Algorithm 2 Abstraction evaluation
Open Source Code Yes All simulations are available online1. 1https://github.com/FMZennaro/Causal Abstraction/tree/main/ papers/2023-quantifying-consistency-and-infoloss
Open Datasets Yes using a lung cancer model from [Guyon et al., 2008]
Dataset Splits No The paper mentions 'Empirical distributions are computed from 10^4 samples; means and standard deviations are computed out of 10 repetitions' but does not specify explicit training, validation, or test dataset splits.
Hardware Specification No The paper describes running empirical simulations but does not provide any specific hardware specifications (e.g., GPU/CPU models, memory) used for these experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, or other libraries with their versions) used in the experiments.
Experiment Setup Yes Empirical distributions are computed from 10^4 samples; means and standard deviations are computed out of 10 repetitions. Two different solutions are learned by minimizing either IC or ILL. Three different solutions are learned by minimizing ISIL with the three assessment sets.