Neurosymbolic LoRA: Why and When to Tune Weights vs. Rewrite Prompts

2026-01-19

AI Summary

This work addresses the challenge of simultaneously injecting factual knowledge and controlling stylistic or alignment attributes during large language model (LLM) adaptation. To this end, the authors propose a neuro-symbolic LoRA framework that dynamically integrates numerical fine-tuning (LoRA) with symbolic prompt editing (TextGrad)—a novel combination achieved through a reward-driven classifier that unifies monitoring signals to selectively trigger either weight updates or prompt rewrites as needed. High-quality, reusable training data are generated by leveraging an external LLM to perform symbolic transformations. Extensive experiments across multiple mainstream LLMs demonstrate that the proposed method significantly outperforms purely numerical or purely symbolic baselines, exhibiting particularly strong adaptability and performance gains on data-scarce tasks such as mathematical reasoning.

Technology Category

• Machine Learning: Neuro-Symbolic Learning
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Symbolic Representations

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Large language models (LLMs) can be adapted either through numerical updates that alter model parameters or symbolic manipulations that work on discrete prompts or logical constraints. While numerical fine-tuning excels at injecting new factual knowledge, symbolic updates offer flexible control of style and alignment without retraining. We introduce a neurosymbolic LoRA framework that dynamically combines these two complementary strategies. Specifically, we present a unified monitoring signal and a reward-based classifier to decide when to employ LoRA for deeper factual reconstruction and when to apply TextGrad for token-level edits. Our approach remains memory-efficient by offloading the symbolic transformations to an external LLM only when needed. Additionally, the refined prompts produced during symbolic editing serve as high-quality, reusable training data, an important benefit in data-scarce domains like mathematical reasoning. Extensive experiments across multiple LLM backbones show that neurosymbolic LoRA consistently outperforms purely numerical or purely symbolic baselines, demonstrating superior adaptability and improved performance. Our findings highlight the value of interleaving numerical and symbolic updates to unlock a new level of versatility in language model fine-tuning.

Problem

• neurosymbolic
• LoRA
• prompt rewriting
• fine-tuning
• large language models

Innovation

• Neurosymbolic LoRA
• parameter-efficient fine-tuning
• symbolic prompting
• TextGrad
• hybrid adaptation