Understanding LLM Reasoning for Abstractive Summarization

TL;DR

This paper systematically evaluates how different reasoning strategies impact the quality and factual accuracy of abstractive summaries generated by Large Language Models, revealing trade-offs and limitations.

Contribution

It provides a comprehensive comparison of reasoning strategies and models for summarization, highlighting their strengths, weaknesses, and the importance of faithful compression.

Findings

Explicit reasoning improves fluency but reduces factual accuracy.

Implicit reasoning enhances factual grounding but may lower fluency.

Increasing reasoning steps does not necessarily improve factual consistency.

Abstract

While the reasoning capabilities of Large Language Models (LLMs) excel in analytical tasks such as mathematics and code generation, their utility for abstractive summarization remains widely assumed but largely unverified. To bridge this gap, we first tailor general reasoning strategies to the summarization domain. We then conduct a systematic, large scale comparative study of 8 reasoning strategies and 3 Large Reasoning Models (LRMs) across 8 diverse datasets, assessing both summary quality and faithfulness. Our findings show that reasoning is not a universal solution and its effectiveness is highly dependent on the specific strategy and context. Specifically, we observe a trade-off between summary quality and factual faithfulness: explicit reasoning strategies tend to improve fluency at the expense of factual grounding, while implicit reasoning in LRMs exhibits the inverse pattern. Furthermore, increasing an LRM's internal reasoning budget does not improve, and can even hurt, factual consistency, suggesting that effective summarization demands faithful compression rather than creative over-thinking.