Can Hallucinations Be Useful? Solving Multi-Hop Questions With SLMs By Chaining System-I/II Reasoning
For researchers working with SLMs on complex reasoning tasks, this work offers a novel strategy that leverages hallucinations as useful signals rather than treating them as errors.
The paper proposes a cognitively-inspired framework for Small Language Models (SLMs) that inverts the typical think-first strategy by first generating a quick answer (System-I) and then using that answer to retrieve evidence for deeper reasoning (System-II). This approach outperforms prior think-first methods on multi-step question-answering benchmarks.
Recently, there has been increased interest in Small Language Models (SLMs), which are fast, show good performance, and have lower hardware demands than large language models (LLMs). However, SLMs hallucinate more frequently than LLMs, impacting their ability to solve complex multi-step reasoning problems as early mistakes cascade to the final response. To address this, existing works think-first followed by iterative retrieval to reduce hallucination. We argue that the think-first strategy is not always necessary as we find that: (i) SLMs are often accurately confident in their initial answer and, (ii) hallucinations can actually be beneficial for honing in on the true answer. As such, we position our work as an inversion of this strategy, i.e., answer first-reason later. We propose a cognitively-inspired framework where the model is first allowed to quickly answer the question (System-I (zero-shot)) and then resorts to deeper thinking (System-II) based on evidence retrieved from a knowledge source using the initial hypothesis. By combining System-I and System-II style thinking, we show that our method can outperform prior work that takes the traditional think-first route on various multi-step question-answering benchmarks.