Traditional attempts to enhance the logical reasoning abilities of language models often rely on supervised fine-tuning, limiting their generalization to new tasks or domains. Large Language Models (LLMs), with their capacity to condense vast knowledge, can effectively tackle many tasks. Yet, our experiments reveal a gap in their performance on logical reasoning benchmarks when compared to state-of-the-art fine-tuning based models. To bridge this gap, we present LogicLLM, a first-of-its-kind, fully self-supervised framework for integrating logical reasoning capabilities into LLMs, and activating them via in-context learning. We apply this to two LLM series, FLAN-T5 and LLaMA, with parameter sizes from 3 billion to 33 billion. LogicLLM demonstrates its effectiveness through successful improvements on two logical reasoning benchmarks (ReClor and LogiQA-v2). Additionally, LogicLLM based on FLAN-T5-11B attains comparable results to ChatGPT, and evaluations with LLaMA-based models on three language understanding benchmarks (RACE, MMLU and Big-Bench-Hard) confirm that the improvements come without compromising the model's general language understanding capabilities.

@inproceedings{Fangkai-et-al-NAACL-24, abstract = {Traditional attempts to enhance the logical reasoning abilities of language models often rely on supervised fine-tuning, limiting their generalization to new tasks or domains. Large Language Models (LLMs), with their capacity to condense vast knowledge, can effectively tackle many tasks. Yet, our experiments reveal a gap in their performance on logical reasoning benchmarks when compared to state-of-the-art fine-tuning based models. To bridge this gap, we present LogicLLM, a first-of-its-kind, fully self-supervised framework for integrating logical reasoning capabilities into LLMs, and activating them via in-context learning. We apply this to two LLM series, FLAN-T5 and LLaMA, with parameter sizes from 3 billion to 33 billion. LogicLLM demonstrates its effectiveness through successful improvements on two logical reasoning benchmarks (ReClor and LogiQA-v2). Additionally, LogicLLM based on FLAN-T5-11B attains comparable results to ChatGPT, and evaluations with LLaMA-based models on three language understanding benchmarks (RACE, MMLU and Big-Bench-Hard) confirm that the improvements come without compromising the model's general language understanding capabilities.}, address = {Mexico City, Mexico}, author = {Fangkai Jiao and Zhiyang Teng and Bosheng Ding and Zhengyuan Liu and Nancy Chen and Shafiq Joty}, booktitle = {2024 Annual Conference of the North American Chapter of the Association for Computational Linguistics}, issue = {}, pages = {}, series = {NAACL-24}, title = {{Exploring Self-supervised Logic-enhanced Training for Large Language Models}}, url = {https://arxiv.org/abs/2305.13718}, year = {2024} }