Combining Constraint Programming Reasoning with Large Language Model Predictions
This work addresses text generation challenges for applications requiring both semantic coherence and structural adherence, though it appears incremental as it builds on existing GenCP and OTFS methods.
The paper tackles the problem of text generation under structural constraints by combining Constraint Programming (CP) for constraints with a Large Language Model (LLM) for meaning, resulting in faster and better performance compared to Beam Search while ensuring all constraints are satisfied.
Constraint Programming (CP) and Machine Learning (ML) face challenges in text generation due to CP's struggle with implementing "meaning'' and ML's difficulty with structural constraints. This paper proposes a solution by combining both approaches and embedding a Large Language Model (LLM) in CP. The LLM handles word generation and meaning, while CP manages structural constraints. This approach builds on GenCP, an improved version of On-the-fly Constraint Programming Search (OTFS) using LLM-generated domains. Compared to Beam Search (BS), a standard NLP method, this combined approach (GenCP with LLM) is faster and produces better results, ensuring all constraints are satisfied. This fusion of CP and ML presents new possibilities for enhancing text generation under constraints.