WILT: A Multi-Turn, Memorization-Robust Inductive Logic Benchmark for LLMs
Keywords: Multi-Turn, Inductive Logic, Hypothesis Space Modeling, Overfitting Robustness, Benchmark
TL;DR: A new multi-turn inductive logic benchmark where each test case starts from the same prompt; reveals sensible ordering of reasoning capability.
Abstract: While large language models (LLMs) have shown impressive abilities across a wide range of domains, they still encounter significant challenges in reasoning tasks that require gathering evidence over multiple turns and drawing logical conclusions from this evidence. These challenges present significant obstacles for LLM chat user interfaces, which rely on multi-turn interactions to facilitate effective collaboration. This limitation leads to real-world issues; for example, service chatbots must gather necessary information from customers over multiple turns to diagnose and resolve problems effectively. Despite the multi-turn nature of many real-world LLM use cases, most existing benchmarks rely on carefully curated single-turn tests, which often blur the line between memorization and genuine reasoning. To address this, we introduce the $\textbf{Wason Inductive Logic Test (WILT)}$, a simple yet challenging multi-turn reasoning benchmark designed to resist memorization. WILT is inspired by the Wason 2-4-6 task, where participants must infer a basic boolean function involving three variables (e.g., $x < y < z$) by proposing test cases (such as $(2, 4, 6)$). In WILT, each test starts from a clean slate, with only the initial instructions provided, preventing models from relying on pre-learned responses. Over several turns, models must interact with the environment by suggesting test cases to narrow the possible hypotheses and ultimately infer the hidden function based on the outcomes. Our findings reveal that LLMs struggle with this task, exhibiting various strengths and weaknesses: some are better at narrowing down the hypothesis space by proposing valuable test cases, while others are more adept at deducing the hidden function from observed cases. Despite these variations, the best-performing model achieves only 28\% accuracy, highlighting a significant gap in LLM performance on complex multi-turn reasoning tasks.
Primary Area: datasets and benchmarks
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 5272
Loading