Prompt Repetition Improves Non-Reasoning LLMs

When not using reasoning, repeating the input prompt improves performance for popular models (Gemini, GPT, Claude, and Deepseek) without increasing the number of generated tokens or latency.

cs.LG cs.AI cs.CL Computer Science

View graph Source

Paper

References (18)

01Asking Again and Again: Exploring LLM Robustness to Repeated QuestionsSagi Shaier, Mario Sanz-Guerrero2024 · arXiv.org · 10 citations In Library

02Selective Attention Improves TransformerYaniv Leviathan, Matan Kalman, Yossi Matias2024 · International Conference on Learning Representations · 31 citations In Library

03Repetition Improves Language Model EmbeddingsJacob Mitchell Springer, Suhas Kotha, Daniel Fried et al.2024 · International Conference on Learning Representations · 77 citations In Library

04Re-Reading Improves Reasoning in Large Language ModelsXiaohan Xu, Chongyang Tao, Tao Shen et al.2023 · Conference on Empirical Methods in Natural Language Processing · 69 citations In Library

05Fast Inference from Transformers via Speculative DecodingYaniv Leviathan, Matan Kalman, Yossi Matias2022 · International Conference on Machine Learning · 1.8k citations In Library

06Chain of Thought Prompting Elicits Reasoning in Large Language ModelsJason Wei, Xuezhi Wang, Dale Schuurmans et al.2022 · Neural Information Processing Systems · 21k citations In Library

07Training Verifiers to Solve Math Word ProblemsK. Cobbe, Vineet Kosaraju, Mo Bavarian et al.2021 · arXiv.org · 9.9k citations In Library

08Measuring Mathematical Problem Solving With the MATH DatasetDan Hendrycks, Collin Burns, Saurav Kadavath et al.2021 · NeurIPS Datasets and Benchmarks · 6.0k citations In Library

09Exploring the Limits of Transfer Learning with a Unified Text-to-Text TransformerColin Raffel, Noam Shazeer, Adam Roberts et al.2019 · Journal of machine learning research · 27k citations In Library

10Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question AnsweringTodor Mihaylov, Peter Clark, Tushar Khot et al.2018 · Conference on Empirical Methods in Natural Language Processing · 2.6k citations In Library

11Think you have Solved Question Answering? Try ARC, the AI2 Reasoning ChallengePeter Clark, Isaac Cowhey, Oren Etzioni et al.2018 · arXiv.org · 5.2k citations In Library

12Note on the sampling error of the difference between correlated proportions or percentagesQ. Mcnemar1947 · Psychometrika · 4.6k citations

Scroll for more · 6 remaining

Similar papers

Asking Again and Again: Exploring LLM Robustness to Repeated QuestionsSagi Shaier, Sanz-Guerrero, Mario, von der Wense, Katharina2024 · arXiv (Cornell University) · 8 citations In Library

Enhancing Reasoning Across GPT, Gemini, and Claude — Time-First Structural Prompting for Large Language ModelsKanazawa, Ryuki2025 · Zenodo (CERN European Organization for Nuclear Research) In Library

Enhancing Reasoning Across GPT, Gemini, and Claude — Time-First Structural Prompting for Large Language ModelsKanazawa, Ryuki2025 · Zenodo (CERN European Organization for Nuclear Research) In Library

Instances Need More Care: Rewriting Prompts for Instances with LLMs in the Loop Yields Better Zero-Shot PerformanceSaurabh Srivastava, Chengyue Huang, Weiguo Fan et al.2023 · Annual Meeting of the Association for Computational Linguistics · 12 citations In Library

Progressive-Hint Prompting Improves Reasoning in Large Language ModelsChuanyang Zheng, Zhengying Liu, Enze Xie et al.2023 · arXiv (Cornell University) · 155 citations In Library

RankPrompt: Step-by-Step Comparisons Make Language Models Better ReasonersChi Hu, Yuan Ge, Xiangnan Ma et al.2024 · International Conference on Language Resources and Evaluation · 10 citations In Library

DNR Bench: Benchmarking Over-Reasoning in Reasoning LLMsOluwanifemi Bamgbose, Masoud Hashemi, Sathwik Tejaswi Madhusudhan et al.2026 · Proceedings of the AAAI Conference on Artificial Intelligence In Library

Fast Thinking with Structured Prompts: Enabling LLM Reasoning Without Chain-of-Thought GenerationKirill Morozov, Liubov Chubarova, Irina Piontkovskaya2025 · International conference Recent advances in natural language processing · 0 citations In Library

Short-Path Prompting in LLMs: Analyzing Reasoning Instability and Solutions for Robust PerformanceZuoli Tang, Junjie Ou, Kaiqin Hu et al.2025 · arXiv.org · 0 citations In Library

Reasoning Models Can Be Effective Without ThinkingWenjie Ma, Jingxuan He, Charlie Snell et al.2025 · arXiv (Cornell University) In Library

Source of record

arXiv

44B

Welcome.

This panel sticks with you. Pick anything below and it opens right beside it, so you can dig through 60,000-plus records without ever losing your spot here.

Travel 44B

LibraryPapers, policy, standards, statute LabsEvery organization building AI ModelsIntelligence, price, and speed BenchmarksThe evaluation catalog IncidentsAI failures and harms SearchOne field across everything DashboardThe state of AI accountability in NY 44B RegistryThe Art. 44-B compliance portal