Symbolic-Neural Soft-Logic Reasoning: Towards Robust and Verifiable Thinking Chains via Cooperative Evolution

2026-05-25

AI Summary

This work addresses the unreliability of chain-of-thought reasoning in large language models (LLMs) caused by neural generation stochasticity, as well as limitations in existing neuro-symbolic approaches—such as hallucinated translations and mismatches between natural language and formal logic. To overcome these challenges, the paper proposes a Symbolic-Neural Soft Reasoning (SSR) framework that innovatively integrates soft logical mechanisms to bridge LLMs with symbolic reasoning. By relaxing strict logical determinism while preserving verifiability, SSR employs a co-evolutionary training strategy to automatically generate human-like, verifiable reasoning chains. This approach substantially enhances the model’s reasoning robustness, interpretability, and intrinsic deductive capacity. Experimental results demonstrate that SSR consistently outperforms state-of-the-art methods across multiple models and benchmarks, and it further enables cross-disciplinary applications, including mathematical reasoning.

Abstract

Large Language Models (LLMs) have demonstrated impressive progress in complex reasoning tasks, largely driven by the Chain-of-Thought (CoT) paradigm, which decomposes difficult problems into intermediate steps. However, CoT reasoning remains fundamentally constrained by the probabilistic nature of neural generation, leading to unfaithful reasoning chains that undermine reliability. Neuro-symbolic approaches attempt to address these issues by combining LLMs with symbolic solvers, yet they face persistent challenges, including hallucinated translations, the mismatch between natural language and formal logic, and the limited enhancement of the LLM's intrinsic reasoning ability. To overcome these limitations, we propose Symbolic-Neural Soft-Logic Reasoning (SSR), a unified framework that integrates LLMs with symbolic reasoning and improves robustness by relaxing strict logical determinism while preserving verifiability. Our approach improves reasoning performance, automatically generates verifiable and human-like logical thinking chains for training and fine-tuning, and facilitates cross-disciplinary applications such as AI for mathematics. Experiments across multiple models and benchmarks demonstrate that SSR consistently outperforms existing reasoning frameworks, highlighting its effectiveness in enhancing both the robustness and interpretability of LLM reasoning.

Problem

• Chain-of-Thought
• Neuro-symbolic reasoning
• LLM reliability
• Logical verifiability
• Reasoning robustness

Innovation

• Symbolic-Neural Integration
• Soft Logic Reasoning
• Verifiable Chain-of-Thought
• Robust LLM Reasoning
• Cooperative Evolution