topo-align / examples /README.md

feat: Introduce `equilib` package with RLHF steering and Sperner dataset generation examples, and add project license.

258dd6d 3 days ago

551 Bytes

Examples

This directory contains standalone scripts demonstrating the capabilities of the Topo-Align (Equilib) library.

rlhf_steering_demo.py: Simulates a Multi-Objective RLHF environment (Helpfulness, Safety, Verbosity) and uses the topological solver to find optimal reward mixing weights.
generate_sperner_dataset.py: The utility used to create the sperner-bench dataset, demonstrating how to generate high-dimensional simplex triangulations.

How to run

From the repository root:

python examples/rlhf_steering_demo.py