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 Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Matching a Desired Causal State via Shift Interventions
Authors: Jiaqi Zhang, Chandler Squires, Caroline Uhler
NeurIPS 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In line with our theoretical results, we also demonstrate experimentally that our proposed active learning strategies require fewer interventions compared to several baselines. [...] We now evaluate our algorithms in several synthetic settings. |
| Researcher Affiliation | Academia | Jiaqi Zhang LIDS, EECS, and IDSS, MIT EMAIL Chandler Squires LIDS, EECS, and IDSS, MIT EMAIL Caroline Uhler LIDS, EECS, and IDSS, MIT EMAIL |
| Pseudocode | Yes | Algorithm 1: Active Learning for Causal Mean Matching |
| Open Source Code | Yes | Code is publicly available at: https://github.com/uhlerlab/causal_mean_matching. |
| Open Datasets | No | The paper describes generating '100 problem instances' in 'synthetic settings' using graph generation models (Erdös-Rényi, Barabási Albert, etc.) but does not specify a publicly available or open dataset that was used for training. |
| Dataset Splits | No | The paper does not explicitly provide training/test/validation dataset splits. It describes generating problem instances for evaluation. |
| Hardware Specification | No | The paper states that experiments were run in 'synthetic settings' but does not provide any specific hardware details such as GPU/CPU models or memory used. |
| Software Dependencies | No | While the paper provides a link to its code repository, it does not explicitly list any software dependencies with specific version numbers within the paper's text. |
| Experiment Setup | Yes | Each setting considers a particular graph type, number of nodes p in the graph and number of perturbation targets |I | p in the matching intervention. We generate 100 problem instances in each setting. [...] The graph size p in our simulations ranges from 10 to 1000, while the number of perturbation targets ranges from 1 to min{p, 100}. [...] Each algorithm is run with sparsity constraint S = 1. [...] Finally, we consider the effect of the sparsity constraint S in Figure 5c with |I | = 50. |