Published August 20, 2025
| Version v1
Software
Open
Artifact for "Active Learning for Neurosymbolic Program Synthesis"
Description
This is the artifact for the paper "Active Learning for Neurosymbolic Program Synthesis."
This artifact supports the experiments presented in our OOPSLA 2025 paper.
Our paper makes the following key contributions:
- We define the neurosymbolic active learning problem and propose the first algorithm for solving it.
- We introduce constrained conformal evaluation (CCE) as a new type of program semantics that takes into account both user feedback and prediction sets obtained through conformal prediction.
- We implement our approach in a new tool called SmartLabel and instantiate it on 3 neurosymbolic domains.
This artifact contains our tool, SmartLabel, applied to the 3 application domains described in the paper, along with all baselines and ablations studied in our evaluation.
Files
SmartLabelArtifact.zip
Additional details
Software
- Repository URL
- https://github.com/celestebarnaby/SmartLabel
- Programming language
- Python