research
NeuroProlog framework combines neural networks with symbolic reasoning to fix LLM math errors
Researchers introduce NeuroProlog, a neurosymbolic framework that compiles math word problems into executable Prolog programs with formal verification guarantees. A multi-task "Cocktail" training strategy achieves significant accuracy improvements on GSM8K: +5.23% on Qwen-32B, +3.43% on GPT-OSS-20B, and +5.54% on Llama-3B compared to single-task baselines.