AlgoPilot: Cross-Paradigm Reasoning in Language Models via Strategy Selection and Guidance

arXiv:2611.04079v1 Announce Type: new Abstract: Real-world problems span diverse domains, including program synthesis, symbolic reasoning, planning, and optimization, and often require fundamentally different solution paradigms. A central challenge for LLM-based reasoning is twofold: identifying an appropriate problem-solving approach and formulating it correctly. In this paper, we propose AlgoPilot, a cross-paradigm reasoning framework that enables adaptive selection of the solving approaches and structured problem formulation prior to execution.