LogicBench: Towards Systematic Evaluation of Logical Reasoning Ability of Large Language Models

Recently developed large language models (LLMs) have been shown to perform remarkably well on a wide range of language understanding tasks. But, can they really "reason" over the natural language? This question has been receiving significant research attention and many reasoning skills such as commonsense, numerical, and qualitative have been studied. However, the crucial skill pertaining to 'logical reasoning' has remained underexplored. Existing work investigating this reasoning ability of LLMs has focused only on a couple of inference rules (such as modus ponens and modus tollens) of propositional and first-order logic. Addressing the above limitation, we comprehensively evaluate the logical reasoning ability of LLMs on 25 different reasoning patterns spanning over propositional, first-order, and non-monotonic logics. To enable systematic evaluation, we introduce LogicBench, a natural language question-answering dataset focusing on the use of a single inference rule. We conduct detailed analysis with a range of LLMs such as GPT-4, ChatGPT, Gemini, Llama-2, and Mistral using chain-of-thought prompting. Experimental results show that existing LLMs do not fare well on LogicBench; especially, they struggle with instances involving complex reasoning and negations. Furthermore, they sometimes overlook contextual information necessary for reasoning to arrive at the correct conclusion. We believe that our work and findings facilitate future research for evaluating and enhancing the logical reasoning ability of LLMs 1 .

Context: Block A and block B are both heavy objects that are typically found on the table. However, there is a possibility that block A might not follow this usual convention. It is important to note this exception.

Context: In a room filled with various objects, two heavy blocks, block A and block B, stand out. Normally, heavy blocks like these are placed on the table, but surprisingly, block A is not found on the table. On the other hand, block B grabs attention with its vibrant red color.

Context: In this situation, there are three heavy blocks: A, B, and C. Typically, heavy blocks are found on the table. However, it is known that at least one of the blocks, either A or B, is not currently on the table.

Context: John confidently states that the vehicle is situated in the driveway, while Sara adamantly counters, asserting that it is not parked inside the garage.

Conclusion: If John's evidence is more reliable than Sara's then the car is parked in the driveway.