# What Matters to an LLM? Behavioral and Computational Evidences from Summarization

arXiv: https://arxiv.org/abs/2602.00459 · HTML: https://mlsanity.com/paper/2602.00459/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 27 of its 55 references that are in the map.
- First posted on arXiv 31 Jan 2026.

## Builds on (most cited first, 27 of 27)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization](https://mlsanity.com/paper/1808.08745/) (2018)
- [Get To The Point: Summarization with Pointer-Generator Networks](https://mlsanity.com/paper/1704.04368/) (2017)
- [Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned](https://mlsanity.com/paper/1905.09418/) (2019)
- [Discovering Latent Knowledge in Language Models Without Supervision](https://mlsanity.com/paper/2212.03827/) (2022)
- [SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization](https://mlsanity.com/paper/1911.12237/) (2019)
- [Dissecting Recall of Factual Associations in Auto-Regressive Language Models](https://mlsanity.com/paper/2304.14767/) (2023)
- [Benchmarking Large Language Models for News Summarization](https://mlsanity.com/paper/2301.13848/) (2023)
- [LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations](https://mlsanity.com/paper/2410.02707/) (2024)
- [Probing the Probing Paradigm: Does Probing Accuracy Entail Task Relevance?](https://mlsanity.com/paper/2005.00719/) (2020)
- [Annotating and Modeling Fine-grained Factuality in Summarization](https://mlsanity.com/paper/2104.04302/) (2021)
- [Content Selection in Deep Learning Models of Summarization](https://mlsanity.com/paper/1810.12343/) (2018)
- [Probing Classifiers are Unreliable for Concept Removal and Detection](https://mlsanity.com/paper/2207.04153/) (2022)
- [Predicting is not Understanding: Recognizing and Addressing Underspecification in Machine Learning](https://mlsanity.com/paper/2207.02598/) (2022)
- [Analyzing LLM Behavior in Dialogue Summarization: Unveiling Circumstantial Hallucination Trends](https://mlsanity.com/paper/2406.03487/) (2024)
- [A Glitch in the Matrix? Locating and Detecting Language Model Grounding with Fakepedia](https://mlsanity.com/paper/2312.02073/) (2023)
- [On Positional Bias of Faithfulness for Long-form Summarization](https://mlsanity.com/paper/2410.23609/) (2024)
- [Bias in News Summarization: Measures, Pitfalls and Corpora](https://mlsanity.com/paper/2309.08047/) (2023)
- [Behavioral Analysis of Information Salience in Large Language Models](https://mlsanity.com/paper/2502.14613/) (2025)
- [The Dead Salmons of AI Interpretability](https://mlsanity.com/paper/2512.18792/) (2025)
- [Demoting the Lead Bias in News Summarization via Alternating Adversarial Learning](https://mlsanity.com/paper/2105.14241/) (2021)
- [Human Guided Exploitation of Interpretable Attention Patterns in Summarization and Topic Segmentation](https://mlsanity.com/paper/2112.05364/) (2021)
- [PSentScore: Evaluating Sentiment Polarity in Dialogue Summarization](https://mlsanity.com/paper/2307.12371/) (2023)
- [Can GPT models Follow Human Summarization Guidelines? A Study for Targeted Communication Goals](https://mlsanity.com/paper/2310.16810/) (2023)
- [Effectiveness of French Language Models on Abstractive Dialogue Summarization Task](https://mlsanity.com/paper/2207.08305/) (2022)
- [QUARTZ : QA-based Unsupervised Abstractive Refinement for Task-oriented Dialogue Summarization](https://mlsanity.com/paper/2509.26302/) (2025)

## Abstract (arXiv)

Large Language Models (LLMs) are now state-of-the-art at summarization, yet the internal notion of importance that drives their information selections remains hidden. We propose to investigate this by combining behavioral and computational analyses. Behaviorally, we generate a series of length-controlled summaries for each document and derive empirical importance distributions based on how often each information unit is selected. These reveal that LLMs converge on consistent importance patterns, sharply different from pre-LLM baselines, and that LLMs cluster more by family than by size. Computationally, we identify that certain attention heads align well with empirical importance distributions, and that middle-to-late layers are strongly predictive of importance. Together, these results provide initial insights into what LLMs prioritize in summarization and how this priority is internally represented, opening a path toward interpreting and ultimately controlling information selection in these models.