Can Hallucinations Be Useful? Solving Multi-Hop Questions With SLMs By Chaining System-I/II Reasoning
Quick Answer
This study proposes a novel approach for Small Language Models (SLMs) to tackle multi-hop questions by answering first and reasoning later, leveraging initial confidence and beneficial hallucinations.
Quick Take
This study proposes a novel approach for Small Language Models (SLMs) to tackle multi-hop questions by answering first and reasoning later, leveraging initial confidence and beneficial hallucinations. The method outperforms traditional think-first strategies on various benchmarks, demonstrating that SLMs can effectively combine quick responses with deeper reasoning.
Key Points
- SLMs often display accurate confidence in initial answers despite higher hallucination rates.
- The proposed method integrates System-I (zero-shot) and System-II reasoning effectively.
- Results show improved performance on multi-step question-answering benchmarks.
- Traditional think-first strategies may not always be necessary for SLMs.
Paper Resources
Article Content
From source RSS / original summaryarXiv:2605. 27596v1 Announce Type: new Abstract: 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.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →Quantifying Prior Dominance in Systems
The study introduces the Normalized Context Utilization (NCU) metric to evaluate Retrieval-Augmented Generation (RAG) systems, revealing that Small Language Models (SLMs) outperform larger models in factual extraction. The findings indicate that traditional scaling laws yield diminishing returns, with a commercial API frequently failing against adversarial evidence due to systemic confidence collapse.