ExploreToM
A framework for generating diverse and challenging psychological theory data on a large scale
0
- Generate story background: Use the story_context_generator.py script to generate the story background.
- Perform A * search: Use the story_structure_searcher.py script to perform A * search to generate complex story structures.
- Fill the generated story: Use the story_structure_inlirer.py script to fill the generated story.
- Statistical analysis: Conduct statistical analysis on the generated data using the compute_statistics.comy script.
- Functional testing: Run tests_belief _ tracker. py and tests_story _ structure _ filler. py for functional testing.
- Model loading: Use VLLM (Large Language Model) to load and run the model.
Product Details
ExploreToM is a framework developed by Facebook Research aimed at generating diverse and challenging psychological theory data on a large scale for enhancing the training and evaluation of large language models (LLMs). This framework utilizes the A * search algorithm to generate complex story structures and novel, diverse, and reasonable scenarios on custom domain specific languages to test the limits of LLMs.
