6B-02
Chain-of-Logic: Symbolic Knowledge Navigates LLM’s Reasoning
○孫 晶鈺，ホンジェイ ジェイ，竹内 亨（NTT研究所）
We iteratively cultivate explicit rules through symbolic reasoning as necessary and channel the structured knowledge back into the LLM prompts to evolve the output for every loop. This enrichment improves the LLM’s capability of multi-hop reasoning with each iteration.
