The paper introduces SymbolicAI, a framework that uses a logic-based approach for concept learning and flow management in generative processes. It integrates generative models with various solvers, treating large language models (LLMs) as semantic parsers that execute tasks based on natural and formal language instructions. The framework uses probabilistic programming principles and different programming paradigms for complex tasks. It allows transitioning between various foundational models with zero-shot learning capabilities and specialized models or solvers for specific problems. The framework also introduces a quality measure called the VERTEX score for evaluating computational graphs.
Publication date: 1 Feb 2024
Project Page: https://github.com/ExtensityAI/symbolicai, https://github.com/ExtensityAI/benchmark
Paper: https://arxiv.org/pdf/2402.00854
