CoChat
Sign Up
Cited by 185Topic ModelingNatural Language Processing TechniquesMultimodal Machine Learning Applications

Same Task, More Tokens: the Impact of Input Length on the Reasoning Performance of Large Language Models

Open access
Levy, Mosh, Jacoby, Alon, Goldberg, Yoav·2024·Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
DOI 10.18653/v1/2024.acl-long.818
Open this paper in CoChat — free accountOpen PDF

Abstract

This paper explores the impact of extending input lengths on the capabilities of Large Language Models (LLMs).Despite LLMs advancements in recent times, their performance consistency across different input lengths is not well understood.We investigate this aspect by introducing a novel QA reasoning framework, specifically designed to assess the impact of input length.We isolate the effect of input length using multiple versions of the same sample, each being extended with padding of different lengths, types and locations.Our findings show a notable degradation in LLMs' reasoning performance at much shorter input lengths than their technical maximum.We show that the degradation trend appears in every version of our dataset, although at different intensities.Additionally, our study reveals that the traditional metric of next word prediction correlates negatively with performance of LLMs' on our reasoning dataset.We analyse our results and identify failure modes that can serve as useful guides for future research, potentially informing strategies to address the limitations observed in LLMs.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.18653/v1/2024.acl-long.818

Cites

0 citation errors · 1 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • The viability of long-context LLMs as long-term memory storage remains largely uncertain given the many engineering hurdles involved and the apparent limitations of long-context LLMs even within current context lengths.

    “However, the viability of this future remains largely uncertain given the many engineering hurdles involved and the apparent limitations of long-context LLMs, even within current context lengths [41, 45, 96, 21].”

    From HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models (2024) · cites this paper as [41]

    The Introduction reports that reasoning accuracy dropped from 0.92 to 0.68 at 3,000 tokens, well below the tested models’ technical maximum; the Discussion likewise says performance fell before maximum input length. This supports the claimed limitations within current context lengths, but the paper does not assess engineering hurdles or the viability of using long-context LLMs as long-term memory storage. (LLM, conf=0.85)

    Experimental — accuracy not yet measured

  • Long-context studies have shown that language models struggle to utilize relevant information within lengthy inputs, with performance often degrading well below the nominal context-window limit.

    “Longcontext studies have shown that language models struggle to utilize relevant information within lengthy inputs, with performance often degrading well below the nominal context-window limit (Liu et al., 2024; Levy et al., 2024; Fraga, 2024; He et al., 2024; Hsieh et al., 2024; Tian et al., 2025).”

    From Weighted Memory Tree: Remembering What Matters for Long-Horizon LLM Agents (2026) · cites this paper as Levy et al., 2024

  • Models can be sensitive to the position of relevant evidence and can exhibit reasoning degradation as input length increases.

    “Models can be sensitive to the position of relevant evidence and can exhibit reasoning degradation as input length increases (Liu et al., 2024; Levy et al., 2024).”

    From Weighted Memory Tree: Remembering What Matters for Long-Horizon LLM Agents (2026) · cites this paper as Levy et al., 2024

  • Long-context studies show that increasing the available context does not ensure reliable use of all included information.

    “Long-context studies show that increasing the available context does not ensure reliable use of all included information.”

    From Weighted Memory Tree: Remembering What Matters for Long-Horizon LLM Agents (2026) · cites this paper as Levy et al., 2024

  • Very long context windows may continue to lead to degraded performance, as has been seen in current models.

    “Third, it possible that such very long context windows will continue to lead to degraded performance, as has been seen in current models [33].”

    From Episodic Memory in AI Agents Poses Risks that Should be Studied and Mitigated (2025) · cites this paper as [33]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

The publisher hasn’t deposited this paper’s reference list.

Research works printed but not linked: 6

  1. Hong Liu, Sang Michael Xie, Zhiyuan Li, and Tengyu Ma. 2023a. Same pre-training loss, better down- stream: Implicit bias matters for language models. In International Conference on Machine Learning, pages 22188-22214.
  2. PMLR.
  3. Yuning Mao, Pengcheng He, Xiaodong Liu, Yelong Shen, Jianfeng Gao, Jiawei Han, and Weizhu Chen. 2020. Generation-augmented retrieval for open- domain question answering. In Annual Meeting of the Association for Computational Linguistics.
  4. Freda Shi, Xinyun Chen, Kanishka Misra, Nathan Scales, David Dohan, Ed H Chi, Nathanael Schärli, and Denny Zhou. 2023. Large language models can be easily distracted by irrelevant context. In Inter- national Conference on Machine Learning, pages 31210-31227.
  5. PMLR.
  6. Jason Weston, Antoine Bordes, Sumit Chopra, Alexan- der M Rush, Bart Van Merriënboer, Armand Joulin, and Tomas Mikolov. 2016. Towards ai-complete question answering: A set of prerequisite toy tasks. In 4th International Conference on Learning Repre- sentations, ICLR 2016.

Cited by

Sorted by citations

Related

Authors
Levy, M., Jacoby, A., & Goldberg, Y.
Affiliations
Bar-Ilan University (IL)
ORCID
Venue
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Publisher
Association for Computational Linguistics
Published
2024
Pages
15339-15353
ISSN
0736-587X (ISSN-L)
DOI
10.18653/v1/2024.acl-long.818
OpenAlex
W4402683949
Cited by
119
References
0
Citations by year
2026: 48 · 2025: 57 · 2024: 14
FWCI
35.12
Type
conference paper
Language
en
Open access
Licence
cc-by
Locations
  • Proceedings of the conference - Association for Computational Linguistics. Meeting (publishedVersion, cc-by)
  • arXiv (Cornell University) (submittedVersion, cc-by)
  • arXiv (Cornell University) (cc-by)
Funders
European Commission (HORIZON2020)
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
  • Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Multimodal Machine Learning Applications
Keywords
large language models, reasoning performance, question answering, padding, next-word prediction, performance degradation, failure modes

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review