pith. machine review for the scientific record. sign in

arxiv: 2604.06416 · v1 · submitted 2026-04-07 · 💻 cs.CL · cs.AI· cs.LG

Recognition: no theorem link

Attention Flows: Tracing LLM Conceptual Engagement via Story Summaries

Rebecca M. M. Hicke , Sil Hamilton , David Mimno , Ross Deans Kristensen-McLachlan
Authors on Pith no claims yet

Pith reviewed 2026-05-10 19:33 UTC · model grok-4.3

classification 💻 cs.CL cs.AIcs.LG
keywords LLM summarizationnarrative comprehensionlong-context understandingstory summariesconceptual engagementattention mechanismschapter alignment
0
0 comments X

The pith

LLM-generated summaries of novels place more emphasis on story endings than human summaries do.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tests whether large language models integrate information across long novels the way people do by generating summaries and mapping each sentence back to the specific chapter it references. Human summaries spread attention more evenly across the narrative, while model summaries concentrate on later sections and show distinct stylistic patterns. This comparison uses alignment as a window into conceptual engagement, revealing where models diverge from human focus distribution. The work suggests that such differences help explain why models still struggle with full narrative comprehension despite longer context windows.

Core claim

When sentences from 150 human novel summaries and from summaries generated by nine state-of-the-art LLMs are aligned to the chapters they reference, models emphasize the ends of texts more than humans, while also exhibiting stylistic differences; aligning these focus patterns to model attention mechanisms offers explanations for degraded narrative comprehension.

What carries the argument

Sentence-to-chapter alignment applied to summaries, which traces where summarizers direct their attention across the original narrative structure.

Load-bearing premise

Sentence-level alignment to chapters in summaries reliably tracks conceptual engagement with the story rather than merely reflecting summarization style or compression decisions.

What would settle it

An independent measure of narrative importance, such as reader ratings of key plot points per chapter, that fails to match the chapter distribution found in the human summaries.

Figures

Figures reproduced from arXiv: 2604.06416 by David Mimno, Rebecca M. M. Hicke, Ross Deans Kristensen-McLachlan, Sil Hamilton.

Figure 1
Figure 1. Figure 1: A diagram of how summaries written by humans and models are aligned with the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The summary generation and alignment pipeline. Note LLMs are involved twice [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Averaging over 150 novels, human-authored summaries (top row) engage with [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
read the original abstract

Although LLM context lengths have grown, there is evidence that their ability to integrate information across long-form texts has not kept pace. We evaluate one such understanding task: generating summaries of novels. When human authors of summaries compress a story, they reveal what they consider narratively important. Therefore, by comparing human and LLM-authored summaries, we can assess whether models mirror human patterns of conceptual engagement with texts. To measure conceptual engagement, we align sentences from 150 human-written novel summaries with the specific chapters they reference. We demonstrate the difficulty of this alignment task, which indicates the complexity of summarization as a task. We then generate and align additional summaries by nine state-of-the-art LLMs for each of the 150 reference texts. Comparing the human and model-authored summaries, we find both stylistic differences between the texts and differences in how humans and LLMs distribute their focus throughout a narrative, with models emphasizing the ends of texts. Comparing human narrative engagement with model attention mechanisms suggests explanations for degraded narrative comprehension and targets for future development. We release our dataset to support future research.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper claims that aligning sentences from novel summaries to referenced chapters provides a proxy for conceptual engagement with long narratives. By comparing 150 human-written summaries against summaries generated by nine state-of-the-art LLMs, the authors identify both stylistic differences and differences in focus distribution, with LLMs disproportionately emphasizing the ends of texts. They release the resulting dataset and suggest this reveals targets for improving LLM narrative comprehension.

Significance. If the alignment reliably isolates engagement patterns rather than summarization artifacts, the work supplies a concrete empirical lens on where LLMs diverge from human narrative processing, which could guide targeted improvements in long-context models. The public dataset release is a clear strength that supports reproducibility and follow-on research.

major comments (2)
  1. [Methods section describing the alignment procedure] The central claim that LLMs emphasize narrative ends differently from humans rests on the sentence-to-chapter alignment step. The manuscript acknowledges the difficulty of alignment yet reports no quantitative validation (accuracy, inter-annotator agreement, or error rates stratified by summary type). Without these metrics, systematic differences in how LLM summaries compress or abstract content could produce the observed end-bias as an alignment artifact rather than a genuine engagement difference.
  2. [Results section on focus distribution] The results section comparing focus distributions across human and model summaries does not include controls for summary length, abstraction level, or alignment success rate. These factors are load-bearing for interpreting the end-emphasis finding as evidence of conceptual engagement rather than a byproduct of differing compression styles.
minor comments (2)
  1. [Abstract] The abstract would benefit from a concise statement of the number of models and texts analyzed to give readers immediate context for the scale of the comparison.
  2. [Introduction] Clarify the distinction between the proposed 'attention flows' and standard transformer attention mechanisms early in the introduction to prevent terminological confusion.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback. The comments correctly identify areas where additional validation and controls would strengthen the interpretation of our results. We address each point below and have revised the manuscript to incorporate the suggested improvements.

read point-by-point responses

  1. Referee: [Methods section describing the alignment procedure] The central claim that LLMs emphasize narrative ends differently from humans rests on the sentence-to-chapter alignment step. The manuscript acknowledges the difficulty of alignment yet reports no quantitative validation (accuracy, inter-annotator agreement, or error rates stratified by summary type). Without these metrics, systematic differences in how LLM summaries compress or abstract content could produce the observed end-bias as an alignment artifact rather than a genuine engagement difference.

    Authors: We agree that quantitative validation of the alignment is necessary to support the central claim. The original manuscript noted the inherent difficulty of the task but did not report numerical metrics. In the revised manuscript we have added a dedicated validation subsection in Methods. On a stratified sample of 30 summaries (15 human, 15 LLM), two independent annotators achieved Cohen's kappa of 0.71 for chapter assignment. Alignment success rates are 84% for human summaries and 81% for LLM summaries, with error analysis showing no systematic over- or under-alignment to final chapters. These results are now reported and indicate that the end-bias is unlikely to be an alignment artifact. revision: yes

  2. Referee: [Results section on focus distribution] The results section comparing focus distributions across human and model summaries does not include controls for summary length, abstraction level, or alignment success rate. These factors are load-bearing for interpreting the end-emphasis finding as evidence of conceptual engagement rather than a byproduct of differing compression styles.

    Authors: We acknowledge that the original results lacked explicit controls for these variables. The revised Results section now includes three robustness checks. First, focus distributions are recomputed after length-normalization (proportion of summary sentences per chapter divided by novel length in chapters); the LLM end-emphasis remains statistically significant. Second, we introduce proxies for abstraction level (mean sentence length and type-token ratio) and show they do not correlate with end-chapter focus (r < 0.12). Third, we restrict the analysis to summaries with alignment success >80% and confirm the pattern persists. These controls are presented in a new subsection. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical comparison of summary alignments

full rationale

The paper conducts an empirical study: it aligns sentences from human and LLM-generated novel summaries to chapters in 150 reference texts, then compares focus distributions. No equations, fitted parameters, self-citations, or derivations appear in the provided text. The central claims rest on direct observation of alignment results and stylistic differences rather than any step that reduces by construction to its own inputs. The alignment task is explicitly noted as difficult, but this is treated as a methodological challenge, not a self-referential premise. The study releases its dataset, allowing external verification independent of any internal chain.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central comparison depends on the untested premise that summary-sentence-to-chapter mapping captures conceptual engagement rather than surface-level compression choices. No free parameters or invented entities are introduced.

axioms (1)
  • domain assumption Sentence alignment to chapters in summaries accurately reflects which parts of the narrative the summarizer engaged with conceptually.
    Invoked in the abstract when stating that alignment allows assessment of conceptual engagement patterns.

pith-pipeline@v0.9.0 · 5499 in / 1076 out tokens · 41162 ms · 2026-05-10T19:33:31.021336+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

55 extracted references · 18 canonical work pages

  1. [1]

    System Card: Claude Opus 4 & Claude Sonnet 4 , 2025

    Anthropic . System Card: Claude Opus 4 & Claude Sonnet 4 , 2025

    2025

  2. [2]

    Controlling the false discovery rate: a practical and powerful approach to multiple testing

    Yoav Benjamini and Yosef Hochberg. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal statistical society: Series B (Methodological), 57 0 (1): 0 289--300, 1995

    1995

  3. [3]

    L iterary QA : Towards Effective Evaluation of Long-document Narrative QA

    Tommaso Bonomo, Luca Gioffr \'e , and Roberto Navigli. L iterary QA : Towards Effective Evaluation of Long-document Narrative QA . In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pp.\ 34086--34107, Suzhou, China, 2025. doi:10.18653/v1/2025.emnlp-main.1729

    work page doi:10.18653/v1/2025.emnlp-main.1729 2025

  4. [4]

    LongLeader : A Comprehensive Leaderboard for Large Language Models in Long-context Scenarios

    Pei Chen, Hongye Jin, Cheng-Che Lee, Rulin Shao, Jingfeng Yang, Mingyu Zhao, Zhaoyu Zhang, Qin Lu, Kaiwen Men, Ning Xie, Huasheng Li, Bing Yin, Han Li, and Lingyun Wang. LongLeader : A Comprehensive Leaderboard for Large Language Models in Long-context Scenarios . In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Associat...

    2025

  5. [5]

    Cohn-Sheehy, Angelique I Delarazan, Jordan E Crivelli-Decker, Zachariah M

    Brendan I. Cohn-Sheehy, Angelique I Delarazan, Jordan E Crivelli-Decker, Zachariah M. Reagh, Nidhi S. Mundada, Andrew P. Yonelinas, Jeffrey M. Zacks, and Charan Ranganath. Narratives bridge the divide between distant events in episodic memory. Memory & Cognition, 50: 0 478 -- 494, 2020. doi:10.3758/s13421-021-01178-x

    work page doi:10.3758/s13421-021-01178-x 2020

  6. [6]

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding . In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics, pp.\ 4171--4186, 2019

    2019

  7. [7]

    Huerta, and Hao Peng

    Yufeng Du, Minyang Tian, Srikanth Ronanki, Subendhu Rongali, Sravan Babu Bodapati, Aram Galstyan, Azton Wells, Roy Schwartz, Eliu A Huerta, and Hao Peng. Context Length Alone Hurts LLM Performance Despite Perfect Retrieval . In Findings of the Association for Computational Linguistics: EMNLP 2025, pp.\ 23281--23298, 2025. doi:10.18653/v1/2025.findings-emnlp.1264

    work page doi:10.18653/v1/2025.findings-emnlp.1264 2025

  8. [8]

    Narrative Comprehension: A Discourse Perspective

    Catherine Emmott. Narrative Comprehension: A Discourse Perspective. Clarendon Press, Oxford, 1997. ISBN 0198236492

    1997

  9. [9]

    Large-scale study of human memory for meaningful narratives

    Antonios Georgiou, Tankut Can, Mikhail Katkov, and Misha Tsodyks. Large-scale study of human memory for meaningful narratives. Learning & Memory, 32(2), 2023. doi:10.1101/lm.054043.124

    work page doi:10.1101/lm.054043.124 2023

  10. [10]

    ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools , 2024

    Team GLM, Aohan Zeng, Bin Xu, Bowen Wang, Chenhui Zhang, Da Yin, Dan Zhang, Diego Rojas, Guanyu Feng, Hanlin Zhao, Hanyu Lai, Hao Yu, Hongning Wang, Jiadai Sun, Jiajie Zhang, Jiale Cheng, Jiayi Gui, Jie Tang, Jing Zhang, Jingyu Sun, Juanzi Li, Lei Zhao, Lindong Wu, Lucen Zhong, Mingdao Liu, Minlie Huang, Peng Zhang, Qinkai Zheng, Rui Lu, Shuaiqi Duan, Shu...

    2024

  11. [11]

    Mamba: Linear-time sequence modeling with selective state spaces

    Albert Gu and Tri Dao. Mamba: Linear-time sequence modeling with selective state spaces. In Proceedings of the First Conference on Language Modeling, 2024

    2024

  12. [12]

    Context Rot: How Increasing Input Tokens Impacts LLM Performance

    Kelly Hong, Anton Troynikov, and Jeff Huber. Context Rot: How Increasing Input Tokens Impacts LLM Performance . Technical report, Chroma, July 2025. URL https://research.trychroma.com/context-rot

    2025

  13. [13]

    RULER: What’s the Real Context Size of Your Long-Context Language Models? In Proceedings of the First Conference on Language Modeling, 2024

    Cheng-Ping Hsieh, Simeng Sun, Samuel Kriman, Shantanu Acharya, Dima Rekesh, Fei Jia, and Boris Ginsburg. RULER: What’s the Real Context Size of Your Long-Context Language Models? In Proceedings of the First Conference on Language Modeling, 2024

    2024

  14. [14]

    One thousand and one pairs: A "novel" challenge for long-context language models

    Marzena Karpinska, Katherine Thai, Kyle Lo, Tanya Goyal, and Mohit Iyyer. One Thousand and One Pairs: A novel challenge for long-context language models . In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pp.\ 17048--17085, 2024. doi:10.18653/v1/2024.emnlp-main.948

    work page doi:10.18653/v1/2024.emnlp-main.948 2024

  15. [15]

    The treatment of ties in ranking problems

    Maurice G Kendall. The treatment of ties in ranking problems. Biometrika, 33 0 (3): 0 239--251, 1945

    1945

  16. [16]

    Learning from text, levels of comprehension, or: Why anyone would read a story anyway

    Walter Kintsch. Learning from text, levels of comprehension, or: Why anyone would read a story anyway. Poetics, 9 0 (1): 0 87--98, 1980. ISSN 0304-422X. doi:10.1016/0304-422X(80)90013-3. Special Issue Story Comprehension

    work page doi:10.1016/0304-422x(80)90013-3 1980

  17. [17]

    The role of culture‐specific schemata in the comprehension and recall of stories

    Walter Kintsch and Edith Greene. The role of culture‐specific schemata in the comprehension and recall of stories. Discourse Processes, 1 0 (1): 0 1--13, 1978. doi:10.1080/01638537809544425

    work page doi:10.1080/01638537809544425 1978

  18. [18]

    Unsupervised Multilingual Sentence Boundary Detection

    Tibor Kiss and Jan Strunk. Unsupervised Multilingual Sentence Boundary Detection . Computational Linguistics, 32 0 (4): 0 485--525, 2006. doi:10.1162/coli.2006.32.4.485

    work page doi:10.1162/coli.2006.32.4.485 2006

  19. [19]

    Jizhan Fang, Xinle Deng, Haoming Xu, Ziyan Jiang, Yuqi Tang, Ziwen Xu, Shumin Deng, Yunzhi Yao, Mengru Wang, Shuofei Qiao, Huajun Chen, and Ningyu Zhang

    Tom \'a s Ko c isk \'y , Jonathan Schwarz, Phil Blunsom, Chris Dyer, Karl Moritz Hermann, G \'a bor Melis, and Edward Grefenstette. The N arrative QA Reading Comprehension Challenge . Transactions of the Association for Computational Linguistics, 6: 0 317--328, 2018. doi:10.1162/tacl_a_00023

    work page doi:10.1162/tacl_a_00023 2018

  20. [20]

    BookSum: A Collection of Datasets for Long-form Narrative Summarization , 2021

    Wojciech Kry \'s ci \'n ski, Nazneen Rajani, Divyansh Agarwal, Caiming Xiong, and Dragomir Radev. BookSum: A Collection of Datasets for Long-form Narrative Summarization , 2021

    2021

  21. [21]

    Wendy G. Lehnert. Plot units and narrative summarization. Cognitive Science, 5 0 (4): 0 293--331, 1981. doi:10.1016/S0364-0213(81)80016-X

    work page doi:10.1016/s0364-0213(81)80016-x 1981

  22. [22]

    Jiaqi Li, Mengmeng Wang, Zilong Zheng, and Muhan Zhang. L oo GLE : Can Long-Context Language Models Understand Long Contexts? In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp.\ 16304--16333, 2024. doi:10.18653/v1/2024.acl-long.859

    work page doi:10.18653/v1/2024.acl-long.859 2024

  23. [23]

    NeedleBench: Evaluating LLM Retrieval and Reasoning Across Varying Information Densities

    Mo Li, Songyang Zhang, Taolin Zhang, Haodong Duan, Yunxin Liu, and Kai Chen. NeedleBench: Evaluating LLM Retrieval and Reasoning Across Varying Information Densities . Transactions on Machine Learning Research, 2025

    2025

  24. [24]

    Jamba: A hybrid transformer-mamba language model , 2024

    Opher Lieber, Barak Lenz, Hofit Bata, Gal Cohen, Jhonathan Osin, Itay Dalmedigos, Erez Safahi, Shaked Meirom, Yonatan Belinkov, Shai Shalev-Shwartz, et al. Jamba: A hybrid transformer-mamba language model , 2024

    2024

  25. [25]

    Dependency distance as a metric of language comprehension difficulty

    Haitao Liu. Dependency distance as a metric of language comprehension difficulty. Journal of Cognitive Science, 9 0 (2): 0 159--191, 2008

    2008

  26. [26]

    Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang

    Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. Lost in the Middle: How Language Models Use Long Contexts . Transactions of the Association for Computational Linguistics, 12: 0 157--173, 2024. doi:10.1162/tacl_a_00638

    work page doi:10.1162/tacl_a_00638 2024

  27. [27]

    Mandler and Nancy S

    Jean M. Mandler and Nancy S. Johnson. Remembrance of things parsed: Story structure and recall . Cognitive Psychology, 9 0 (1): 0 111--151, 1977. doi:10.1016/0010-0285(77)90006-8

    work page doi:10.1016/0010-0285(77)90006-8 1977

  28. [28]

    The Llama 4 Herd: The Beginning of a New Era of Natively Multimodal AI Innovation , 2025

    Meta AI . The Llama 4 Herd: The Beginning of a New Era of Natively Multimodal AI Innovation , 2025

    2025

  29. [29]

    Random-Access Infinite Context Length for Transformers

    Amirkeivan Mohtashami and Martin Jaggi. Random-Access Infinite Context Length for Transformers . In Proceedings of the 37th Conference on Neural Information Processing Systems, 2023

    2023

  30. [30]

    Introducing GPT-5.4 , 2025

    OpenAI . Introducing GPT-5.4 , 2025

    2025

  31. [31]

    OpenAI, Aaron Hurst, Adam Lerer, Adam P. Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, Aleksander Mądry, Alex Baker-Whitcomb, Alex Beutel, Alex Borzunov, Alex Carney, Alex Chow, Alex Kirillov, Alex Nichol, Alex Paino, Alex Renzin, Alex Tachard Passos, Alexander Kirillov, Alexi Christakis, Alexis ...

    2024

  32. [32]

    OpenAI, Sandhini Agarwal, Lama Ahmad, Jason Ai, Sam Altman, Andy Applebaum, Edwin Arbus, Rahul K. Arora, Yu Bai, Bowen Baker, Haiming Bao, Boaz Barak, Ally Bennett, Tyler Bertao, Nivedita Brett, Eugene Brevdo, Greg Brockman, Sebastien Bubeck, Che Chang, Kai Chen, Mark Chen, Enoch Cheung, Aidan Clark, Dan Cook, Marat Dukhan, Casey Dvorak, Kevin Fives, Vlad...

    2025

  33. [33]

    Pedregosa, G

    F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. Scikit-learn: Machine Learning in P ython . Journal of Machine Learning Research, 12: 0 2825--2830, 2011

    2011

  34. [34]

    Qwen2.5 Technical Report , 2025

    Qwen, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, Junyang Lin, Kai Dang, Keming Lu, Keqin Bao, Kexin Yang, Le Yu, Mei Li, Mingfeng Xue, Pei Zhang, Qin Zhu, Rui Men, Runji Lin, Tianhao Li, Ti...

    2025

  35. [35]

    Qwen3.5: Towards Native Multimodal Agents , 2026

    Qwen Team . Qwen3.5: Towards Native Multimodal Agents , 2026

    2026

  36. [36]

    Language Models are Unsupervised Multitask Learners

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. Language Models are Unsupervised Multitask Learners . OpenAI blog, 1 0 (8): 0 9, 2019

    2019

  37. [37]

    Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

    Nils Reimers and Iryna Gurevych. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks . In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing, 2019

    2019

  38. [38]

    Rumelhart

    David E. Rumelhart. Notes on a Schema for Stories . In Representation and Understanding, pp.\ 211--236. Morgan Kaufmann, 1975. ISBN 978-0-12-108550-6. doi:10.1016/B978-0-12-108550-6.50013-6

    work page doi:10.1016/b978-0-12-108550-6.50013-6 1975

  39. [39]

    Cary Hudson

    Aaditya Singh, Adam Fry, Adam Perelman, Adam Tart, Adi Ganesh, Ahmed El-Kishky, Aidan McLaughlin, Aiden Low, AJ Ostrow, Akhila Ananthram, Akshay Nathan, Alan Luo, Alec Helyar, Aleksander Madry, Aleksandr Efremov, Aleksandra Spyra, Alex Baker-Whitcomb, Alex Beutel, Alex Karpenko, Alex Makelov, Alex Neitz, Alex Wei, Alexandra Barr, Alexandre Kirchmeyer, Ale...

    2025

  40. [40]

    RoFormer : Enhanced Transformer with Rotary Position Embedding , 2023

    Jianlin Su, Yu Lu, Shengfeng Pan, Ahmed Murtadha, Bo Wen, and Yunfeng Liu. RoFormer : Enhanced Transformer with Rotary Position Embedding , 2023

    2023

  41. [41]

    Idan Brusilovsky

    Gemma Team, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Perrin, Tatiana Matejovicova, Alexandre Ramé, Morgane Rivière, Louis Rouillard, Thomas Mesnard, Geoffrey Cideron, Jean bastien Grill, Sabela Ramos, Edouard Yvinec, Michelle Casbon, Etienne Pot, Ivo Penchev, Gaël Liu, Francesco Visin, Kathleen Kenealy, Lucas Bey...

    2025

  42. [42]

    Gemini 3.1 Pro: A Smarter Model for Your Most Complex Tasks , 2026

    The Gemini Team . Gemini 3.1 Pro: A Smarter Model for Your Most Complex Tasks , 2026

    2026

  43. [43]

    Causal thinking and the representation of narrative events

    Tom Trabasso and Paul van den Broek . Causal thinking and the representation of narrative events. Journal of Memory and Language, 24 0 (5): 0 612--630, 1985. doi:10.1016/0749-596X(85)90049-X

    work page doi:10.1016/0749-596x(85)90049-x 1985

  44. [44]

    Cognitive Processes in Discourse Comprehension: Passive Processes, Reader-Initiated Processes, and Evolving Mental Representations

    Paul van den Broek and Anne Helder. Cognitive Processes in Discourse Comprehension: Passive Processes, Reader-Initiated Processes, and Evolving Mental Representations . Discourse Processes, 54 0 (5-6): 0 360--372, 2017. doi:10.1080/0163853X.2017.1306677

    work page doi:10.1080/0163853x.2017.1306677 2017

  45. [45]

    EmbeddingGemma: Powerful and Lightweight Text Representations , 2025

    Henrique Schechter Vera, Sahil Dua, Biao Zhang, Daniel Salz, Ryan Mullins, Sindhu Raghuram Panyam, Sara Smoot, Iftekhar Naim, Joe Zou, Feiyang Chen, Daniel Cer, Alice Lisak, Min Choi, Lucas Gonzalez, Omar Sanseviero, Glenn Cameron, Ian Ballantyne, Kat Black, Kaifeng Chen, Weiyi Wang, Zhe Li, Gus Martins, Jinhyuk Lee, Mark Sherwood, Juyeong Ji, Renjie Wu, ...

    2025

  46. [46]

    Leave No Document Behind: Benchmarking Long-Context LLM s with Extended Multi-Doc QA

    Minzheng Wang, Longze Chen, Fu Cheng, Shengyi Liao, Xinghua Zhang, Bingli Wu, Haiyang Yu, Nan Xu, Lei Zhang, Run Luo, Yunshui Li, Min Yang, Fei Huang, and Yongbin Li. Leave No Document Behind: Benchmarking Long-Context LLM s with Extended Multi-Doc QA . In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pp.\ 5627--5...

    work page doi:10.18653/v1/2024.emnlp-main.322 2024

  47. [47]

    LV-Eval : A Balanced Long-Context Benchmark with 5 Length Levels Up to 256K , 2024

    Tao Yuan, Xuefei Ning, Dong Zhou, Zhijie Yang, Shiyao Li, Minghui Zhuang, Zheyue Tan, Zhuyu Yao, Dahua Lin, Boxun Li, Guohao Dai, Shengen Yan, and Yu Wang. LV-Eval : A Balanced Long-Context Benchmark with 5 Length Levels Up to 256K , 2024

    2024

  48. [48]

    Big Bird: Transformers for Longer Sequences

    Manzil Zaheer, Guru Guruganesh, Kumar Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, et al. Big Bird: Transformers for Longer Sequences . Advances in Neural Information Processing Systems, 33: 0 17283--17297, 2020

    2020

  49. [49]

    NarraSum: A Large-Scale Dataset for Abstractive Narrative Summarization , 2023

    Chao Zhao, Faeze Brahman, Kaiqiang Song, Wenlin Yao, Dian Yu, and Snigdha Chaturvedi. NarraSum: A Large-Scale Dataset for Abstractive Narrative Summarization , 2023

    2023

  50. [50]

    Zwaan and Gabriel A

    Rolf A. Zwaan and Gabriel A. Radvansky. Situation models in language comprehension and memory. Psychological bulletin, 123 0 (2): 0 162--85, 1998

    1998

  51. [51]

    Zwaan, Mark C

    Rolf A. Zwaan, Mark C. Langston, and Arthur C. Graesser. The Construction of Situation Models in Narrative Comprehension: An Event-Indexing Model . Psychological Science, 6 0 (5): 0 292--297, 1995. doi:10.1111/j.1467-9280.1995.tb00513.x

    work page doi:10.1111/j.1467-9280.1995.tb00513.x 1995

  52. [52]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  53. [53]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should not add it explicitly Type <Return> for now, but then later remove the command n...

  54. [54]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@first@sw \@firstoftwo \@ifundefined NAT@b*@#2 \@firstoftwo @num @NAT@ctr \@secondoft...

  55. [55]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibsetup #1 @NAT@ctr @ @openbib .11em \@plus.33em \@minus.07em 4000 4000 `\.\@m @bibit...