# Hallucinate at the Last in Long Response Generation: A Case Study on Long Document Summarization

arXiv: https://arxiv.org/abs/2505.15291 · HTML: https://mlsanity.com/paper/2505.15291/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 45 of its 59 references that are in the map.
- First posted on arXiv 21 May 2025.

## Builds on (most cited first, 30 of 45)

- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [A Survey on Large Language Model based Autonomous Agents](https://mlsanity.com/paper/2308.11432/) (2023)
- [Gemma 3 Technical Report](https://mlsanity.com/paper/2503.19786/) (2025)
- [Generating Long Sequences with Sparse Transformers](https://mlsanity.com/paper/1904.10509/) (2019)
- [Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization](https://mlsanity.com/paper/1808.08745/) (2018)
- [FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation](https://mlsanity.com/paper/2305.14251/) (2023)
- [ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools](https://mlsanity.com/paper/2406.12793/) (2024)
- [Teaching Machines to Read and Comprehend](https://mlsanity.com/paper/1506.03340/) (2015)
- [A Discourse-Aware Attention Model for Abstractive Summarization of Long Documents](https://mlsanity.com/paper/1804.05685/) (2018)
- [Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuning](https://mlsanity.com/paper/2306.14565/) (2023)
- [MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention](https://mlsanity.com/paper/2407.02490/) (2024)
- [Efficient Attentions for Long Document Summarization](https://mlsanity.com/paper/2104.02112/) (2021)
- [Trusting Your Evidence: Hallucinate Less with Context-aware Decoding](https://mlsanity.com/paper/2305.14739/) (2023)
- [MiniCheck: Efficient Fact-Checking of LLMs on Grounding Documents](https://mlsanity.com/paper/2404.10774/) (2024)
- [LongWriter: Unleashing 10,000+ Word Generation from Long Context LLMs](https://mlsanity.com/paper/2408.07055/) (2024)
- [Lookback Lens: Detecting and Mitigating Contextual Hallucinations in Large Language Models Using Only Attention Maps](https://mlsanity.com/paper/2407.07071/) (2024)
- [BooookScore: A systematic exploration of book-length summarization in the era of LLMs](https://mlsanity.com/paper/2310.00785/) (2023)
- [RLAIF-V: Open-Source AI Feedback Leads to Super GPT-4V Trustworthiness](https://mlsanity.com/paper/2405.17220/) (2024)
- [A Thorough Examination of Decoding Methods in the Era of LLMs](https://mlsanity.com/paper/2402.06925/) (2024)
- [Understanding Factual Errors in Summarization: Errors, Summarizers, Datasets, Error Detectors](https://mlsanity.com/paper/2205.12854/) (2022)
- [Abstractive Summarization of Reddit Posts with Multi-level Memory Networks](https://mlsanity.com/paper/1811.00783/) (2018)
- [Found in the Middle: Calibrating Positional Attention Bias Improves Long Context Utilization](https://mlsanity.com/paper/2406.16008/) (2024)
- [Tell Your Model Where to Attend: Post-hoc Attention Steering for LLMs](https://mlsanity.com/paper/2311.02262/) (2023)
- [FineSurE: Fine-grained Summarization Evaluation using LLMs](https://mlsanity.com/paper/2407.00908/) (2024)
- [Factually Consistent Summarization via Reinforcement Learning with Textual Entailment Feedback](https://mlsanity.com/paper/2306.00186/) (2023)
- [FABLES: Evaluating faithfulness and content selection in book-length summarization](https://mlsanity.com/paper/2404.01261/) (2024)
- [LongGenBench: Benchmarking Long-Form Generation in Long Context LLMs](https://mlsanity.com/paper/2409.02076/) (2024)
- [AdaCAD: Adaptively Decoding to Balance Conflicts between Contextual and Parametric Knowledge](https://mlsanity.com/paper/2409.07394/) (2024)

## Built on by (most cited first, 6 of 6)

- [The Curious Case of Factual (Mis)Alignment between LLMs' Short- and Long-Form Answers](https://mlsanity.com/paper/2510.11218/) (2025)
- [From 'May' to 'Is': Certainty Distortion in Language Model Rewriting](https://mlsanity.com/paper/2606.07951/) (2026)
- [On-Device Deep Research at 4B: Exposure Bounds Faithfulness, Retrieval Bounds Coverage](https://mlsanity.com/paper/2607.12257/) (2026)
- [Reliability Scales Inversely: Hallucinations Snowball Faster in Bigger Language Models](https://mlsanity.com/paper/2607.18292/) (2026)
- [Doc2FRC: Length-Consistent Document-Level Machine Translation via Fixed-Range Chunking](https://mlsanity.com/paper/2609.12674/) (2026)
- [From Utterances to Networks: Modelling Slang Adoption and Diffusion Across Subreddits](https://mlsanity.com/paper/2609.25669/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) have significantly advanced text generation capabilities, including tasks like summarization, often producing coherent and fluent outputs. However, faithfulness to source material remains a significant challenge due to the generation of hallucinations. While extensive research focuses on detecting and reducing these inaccuracies, less attention has been paid to the positional distribution of hallucination within generated text, particularly in long outputs. In this work, we investigate where hallucinations occur in LLM-based long response generation, using long document summarization as a key case study. Focusing on the challenging setting of long context-aware long response generation, we find a consistent and concerning phenomenon: hallucinations tend to concentrate disproportionately in the latter parts of the generated long response. To understand this bias, we explore potential contributing factors related to the dynamics of attention and decoding over long sequences. Furthermore, we investigate methods to mitigate this positional hallucination, aiming to improve faithfulness specifically in the concluding segments of long outputs.