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..
Learning Libraries of Subroutines for Neurally–Guided Bayesian Program Induction
Authors: Kevin Ellis, Lucas Morales, Mathias Sablé-Meyer, Armando Solar-Lezama, Josh Tenenbaum
NeurIPS 2018 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We use our model to synthesize functions on lists, edit text, and solve symbolic regression problems, showing how the model learns a domain-specific library of program components for expressing solutions to problems in the domain. ... We evaluated EC2 on random 50/50 test/train split. ... Tbl. 3 compares our model against these alternatives. ... Figure 4: Learning curves for EC2 both with (in orange) and without (in teal) the recognition model. |
| Researcher Affiliation | Academia | Kevin Ellis MIT EMAIL Lucas Morales MIT EMAIL Mathias Sablé-Meyer ENS Paris-Saclay EMAIL Armando Solar-Lezama MIT EMAIL Joshua B. Tenenbaum MIT EMAIL |
| Pseudocode | Yes | Algorithm 1 The EC2 Algorithm |
| Open Source Code | No | The paper does not contain any statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | We tested, but did not train, on the 108 text editing problems from the Sy Gu S [32] program synthesis competition. [32] Rajeev Alur, Dana Fisman, Rishabh Singh, and Armando Solar-Lezama. Sygus-comp 2016: results and analysis. ar Xiv preprint ar Xiv:1611.07627, 2016. |
| Dataset Splits | Yes | We evaluated EC2 on random 50/50 test/train split. ... We tested, but did not train, on the 108 text editing problems from the Sy Gu S [32] program synthesis competition. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU models, or cloud computing instance types used for experiments. |
| Software Dependencies | No | The paper mentions using a 'bidirectional GRU [30]' but does not specify any software names with version numbers for reproducibility. |
| Experiment Setup | Yes | Input: Initial DSL D, set of tasks X, iterations I Hyperparameters: Enumeration timeout T Initialize θ uniform. ... After three iterations, it assembles a DSL... |