On the Brittle Foundations of ReAct Prompting for Agentic Large Language Models

Ranked by how close their title and abstract are to this paper’s. Select some to read them together.

Closest papers

1. Very close Same question Cited by this paper Not relevant

   ReAct: Synergizing Reasoning and Acting in Language Models

   Shunyu Yao, Jeffrey Zhao, Dian Yu et al. · 2022 · International Conference on Learning Representations · cited 13,359 times · Not summarized yet

   ReAct proposes interleaved reasoning and action for sequential decision-making, which this paper tests and finds contributes little to performance.

   Abstract

   While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it to interface with external sources, such as knowledge bases or environments, to gather additional in…

2. Very close Another approach to the problem Not relevant

   Enhancing LLM Problem Solving with REAP: Reflection, Explicit Problem Deconstruction, and Advanced Prompting

   Ryan Lingo, Martin Arroyo, Rajeev Chhajer · 2024 · arXiv.org · preprint, not peer reviewed · cited 6 times · Not summarized yet

   REAP guides problem solving through reflection, decomposition, and generated context, rather than testing ReAct’s reasoning traces and task-example similarity.

   Abstract

   Large Language Models (LLMs) have transformed natural language processing, yet improving their problem-solving capabilities, particularly for complex, reasoning-intensive tasks, remains a persistent challenge. This paper introduces the REAP (Reflection, Explicit Problem Deconstruction, and Advanced Prompting) method, an innovative approach within the dynamic context generation framework. REAP guides LLMs through reflection on the query, deconstructing it into manageable components, and generating relevant context to enhance the solution process. We evaluated REAP using a dataset designed to expose LLM limitations, comparing zero-shot prompting with REAP-enhanced prompts across six state-of-t…

3. Very close Another approach to the problem Not relevant

   ReActIn: Infusing Human Feedback into Intermediate Prompting Steps of Large Language Model

   Manuel Delaflor Rodrguez, Carlos Toxtli, Claire Gendron et al. · 2024 · Human Interaction and Emerging Technologies (IHIET-AI 2024): Artificial Intelligence and Future Applications · cited 2 times · Not summarized yet

   ReActIn adds human feedback to intermediate prompting steps and evaluates its effects on reasoning and decisions, unlike this paper’s sensitivity analysis of ReAct.

   Abstract

   This paper introduces ReActIn, a framework designed to infuse human feedback into the intermediate prompting steps of large language models. The practicality and effectiveness of ReActIn are validated through experiments that apply four established prompting strategies, evaluated both with and without human feedback integration. The proposed architecture's performance is compared against traditional large language models across various tasks using four standard evaluation tests. Our findings reveal that the integration of human feedback has a direct impact on the reasoning, action prompting, and overall decision-making capabilities of the language models. This study underscores the potential…

4. Very close Another approach to the problem Not relevant

   Logic-of-Thought: Injecting Logic into Contexts for Full Reasoning in Large Language Models

   Tongxuan Liu, Wenjiang Xu, Weizhe Huang et al. · 2024 · Computation and Language · cited 40 times · AI summary available

   Logic-of-Thought injects propositional logic into contexts to improve logical reasoning, rather than examining whether ReAct gains come from exemplar similarity.

   Abstract

   Large Language Models (LLMs) have demonstrated remarkable capabilities across various tasks but their performance in complex logical reasoning tasks remains unsatisfactory. Although some prompting methods, such as Chain-of-Thought, can improve the reasoning ability of LLMs to some extent, they suffer from an unfaithful issue where derived conclusions may not align with the generated reasoning chain. To address this issue, some studies employ the approach of propositional logic to further enhance logical reasoning abilities of LLMs. However, the potential omissions in the extraction of logical expressions in these methods can cause information loss in the logical reasoning process, thereby ge…

5. Very close Another approach to the problem Cited by this paper A method it uses Not relevant

   ReST meets ReAct: Self-Improvement for Multi-Step Reasoning LLM Agent

   Renat Aksitov, Sobhan Miryoosefi, Zonglin Li et al. · 2023 · Computation and Language · preprint, not peer reviewed · cited 95 times · AI summary available

   ReST meets ReAct trains an agent on prior reasoning-and-action trajectories, while this paper questions whether ReAct’s prompting itself improves reasoning.

   Abstract

   Answering complex natural language questions often necessitates multi-step reasoning and integrating external information. Several systems have combined knowledge retrieval with a large language model (LLM) to answer such questions. These systems, however, suffer from various failure cases, and we cannot directly train them end-to-end to fix such failures, as interaction with external knowledge is non-differentiable. To address these deficiencies, we define a ReAct-style LLM agent with the ability to reason and act upon external knowledge. We further refine the agent through a ReST-like method that iteratively trains on previous trajectories, employing growing-batch reinforcement learning wi…

6. Very close Another approach to the problem Not relevant

   Meta Reasoning for Large Language Models

   Pei-Zhong Gao, Ao Xie, Shaoguang Mao et al. · 2024 · arXiv.org · preprint, not peer reviewed · cited 27 times · Not summarized yet

   Meta-Reasoning Prompting selects reasoning methods according to task requirements, whereas this paper investigates the source of ReAct’s performance.

   Abstract

   We introduce Meta-Reasoning Prompting (MRP), a novel and efficient system prompting method for large language models (LLMs) inspired by human meta-reasoning. Traditional in-context learning-based reasoning techniques, such as Tree-of-Thoughts, show promise but lack consistent state-of-the-art performance across diverse tasks due to their specialized nature. MRP addresses this limitation by guiding LLMs to dynamically select and apply different reasoning methods based on the specific requirements of each task, optimizing both performance and computational efficiency. With MRP, LLM reasoning operates in two phases. Initially, the LLM identifies the most appropriate reasoning method using task …

7. Very close Another approach to the problem Not relevant

   Take a Step Back: Evoking Reasoning via Abstraction in Large Language Models

   Huaixiu Steven Zheng, Swaroop Mishra, Xinyun Chen et al. · 2023 · Machine Learning · cited 261 times · AI summary available

   Step-Back Prompting uses abstraction to guide reasoning on challenging tasks, rather than testing ReAct’s interleaved traces and examples.

   Abstract

   We present Step-Back Prompting, a simple prompting technique that enables LLMs to do abstractions to derive high-level concepts and first principles from instances containing specific details. Using the concepts and principles to guide the reasoning steps, LLMs significantly improve their abilities in following a correct reasoning path towards the solution. We conduct experiments of Step-Back Prompting with PaLM-2L models and observe substantial performance gains on a wide range of challenging reasoning-intensive tasks including STEM, Knowledge QA, and Multi-Hop Reasoning. For instance, Step-Back Prompting improves PaLM-2L performance on MMLU Physics and Chemistry by 7% and 11%, TimeQA by 27…

8. Very close Another approach to the problem Not relevant

   Rephrase and Respond: Let Large Language Models Ask Better Questions for Themselves

   Yihe Deng, Weitong Zhang, Zixiang Chen et al. · 2023 · Computation and Language · preprint, not peer reviewed · cited 170 times · AI summary available

   Rephrase and Respond has models reformulate questions before answering, while this paper examines task-example similarity in ReAct prompts.

   Abstract

   Misunderstandings arise not only in interpersonal communication but also between humans and Large Language Models (LLMs). Such discrepancies can make LLMs interpret seemingly unambiguous questions in unexpected ways, yielding incorrect responses. While it is widely acknowledged that the quality of a prompt, such as a question, significantly impacts the quality of the response provided by LLMs, a systematic method for crafting questions that LLMs can better comprehend is still underdeveloped. In this paper, we present a method named `Rephrase and Respond' (RaR), which allows LLMs to rephrase and expand questions posed by humans and provide responses in a single prompt. This approach serves as…

Prior work it builds on

1. Close Groundwork for this paper A method it uses Not relevant

   Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

   Jason Wei, Xuezhi Wang, Dale Schuurmans et al. · 2022 · Computation and Language · cited 23,501 times · AI summary available

   Chain-of-Thought Prompting reports gains from reasoning-trace exemplars, a prompting claim relevant to this paper’s tests of ReAct traces.

   Abstract

   We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just…

2. Close Broader context A method it uses Not relevant

   LLMs Can't Plan, But Can Help Planning in LLM-Modulo Frameworks

   Subbarao Kambhampati, Karthik Valmeekam, L. Guan et al. · 2024 · arXiv.org · preprint, not peer reviewed · cited 245 times · Not summarized yet

   LLMs Can't Plan argues that language models cannot plan alone, situating this paper’s challenge to claims of agentic reasoning.

   Abstract

   There is considerable confusion about the role of Large Language Models (LLMs) in planning and reasoning tasks. On one side are over-optimistic claims that LLMs can indeed do these tasks with just the right prompting or self-verification strategies. On the other side are perhaps over-pessimistic claims that all that LLMs are good for in planning/reasoning tasks are as mere translators of the problem specification from one syntactic format to another, and ship the problem off to external symbolic solvers. In this position paper, we take the view that both these extremes are misguided. We argue that auto-regressive LLMs cannot, by themselves, do planning or self-verification (which is after al…

3. Close Broader context Not relevant

   On the Self-Verification Limitations of Large Language Models on Reasoning and Planning Tasks

   Kaya Stechly, Karthik Valmeekam, Subbarao Kambhampati · 2024 · International Conference on Learning Representations · cited 188 times · Not summarized yet

   On the Self-Verification Limitations examines LLM reasoning and planning limits, sharing this paper’s skepticism about claimed reasoning abilities.

   Abstract

   There has been considerable divergence of opinion on the reasoning abilities of Large Language Models (LLMs). While the initial optimism that reasoning might emerge automatically with scale has been tempered thanks to a slew of counterexamples--ranging from multiplication to simple planning--there persists a wide spread belief that LLMs can self-critique and improve their own solutions in an iterative fashion. This belief seemingly rests on the assumption that verification of correctness should be easier than generation--a rather classical argument from computational complexity--which should be irrelevant to LLMs to the extent that what they are doing is approximate retrieval. In this paper,…

4. Close Review A method it uses Not relevant

   A Systematic Survey of Prompt Engineering in Large Language Models: Techniques and Applications

   Pranab Sahoo, Ayush Kumar Singh, Sriparna Saha et al. · 2024 · arXiv.org · preprint, not peer reviewed · cited 1,095 times · AI summary available

   A Systematic Survey reviews prompt-engineering techniques and applications, providing a broader account of the prompting methods examined here.

   Abstract

   Prompt engineering has emerged as an indispensable technique for extending the capabilities of large language models (LLMs) and vision-language models (VLMs). This approach leverages task-specific instructions, known as prompts, to enhance model efficacy without modifying the core model parameters. Rather than updating the model parameters, prompts allow seamless integration of pre-trained models into downstream tasks by eliciting desired model behaviors solely based on the given prompt. Prompts can be natural language instructions that provide context to guide the model or learned vector representations that activate relevant knowledge. This burgeoning field has enabled success across vario…

5. Close Another approach to the problem A method it uses Not relevant

   PreAct: Prediction Enhances Agent's Planning Ability

   Dayuan Fu, Jianzhao Huang, Si-Yuan Lu et al. · 2024 · International Conference on Computational Linguistics · cited 20 times · Not summarized yet

   PreAct adds prediction before reasoning and action and reports better task completion than ReAct, unlike this paper’s analysis of ReAct’s purported gains.

   Abstract

   Addressing the disparity between forecasts and actual results can enable individuals to expand their thought processes and stimulate self-reflection, thus promoting accurate planning. In this research, we present **PreAct**, an agent framework that integrates **pre**diction, **rea**soning, and **act**ion. By utilizing the information derived from predictions, the large language model (LLM) agent can provide a wider range and more strategically focused reasoning. This leads to more efficient actions that aid the agent in accomplishing intricate tasks. Our experimental results show that PreAct surpasses the ReAct method in completing complex tasks and that PreAct's performance can be further i…

Papers that cite it

1. Close Builds on this paper Not relevant

   Structured Prompting for Improving Reasoning Accuracy and Reducing Hallucination in Large Language Models

   Vartika Kesarwani, Chanchal Garg, Prem Kumar et al. · 2026 · International Conference on Computing for Sustainable Global Development · Not summarized yet

   Structured Prompting cites this paper while comparing guided and framed prompts for reasoning accuracy and hallucination reduction.

   Abstract

   In this exploration, it is suggested to develop a robust framework of prompt optimisation to realise better accuracy of reasoning and reduce hallucinations in high-precision natural language processing tasks using large language models, without modification of the model or its retraining. Hallucinations are defined as the wrong or indisputed outputs or the model. This framework compares framed and guided prompts with the baseline prompts and assesses them based on criteria e.g., clarity of task, contextualisation, incremental reasoning and cues of verification. An error of judgment, consistency, and frequency of hallucination are employed in measuring performance. Besides this, an expert-bas…

2. Close Builds on this paper Not relevant

   Exploring Prompt Engineering: A Systematic Review with SWOT Analysis

   Aditi Singh, Abul Ehtesham, Gaurav Kumar Gupta et al. · 2024 · arXiv.org · preprint, not peer reviewed · cited 10 times · Not summarized yet

   Exploring Prompt Engineering cites this paper in its SWOT analysis of prompting techniques and their challenges.

   Abstract

   In this paper, we conduct a comprehensive SWOT analysis of prompt engineering techniques within the realm of Large Language Models (LLMs). Emphasizing linguistic principles, we examine various techniques to identify their strengths, weaknesses, opportunities, and threats. Our findings provide insights into enhancing AI interactions and improving language model comprehension of human prompts. The analysis covers techniques including template-based approaches and fine-tuning, addressing the problems and challenges associated with each. The conclusion offers future research directions aimed at advancing the effectiveness of prompt engineering in optimizing human-machine communication.

3. Close Broader context Not relevant

   LLMs Can't Play Hangman: On the Necessity of a Private Working Memory for Language Agents

   Davide Baldelli, Alipanah Parviz, A. Zouaq et al. · 2026 · arXiv.org · preprint, not peer reviewed · cited 5 times · Not summarized yet

   LLMs Can't Play Hangman studies hidden-state reliability in language agents, a neighboring question about agent limitations rather than ReAct prompting.

   Abstract

   As LLMs move from text completion toward autonomous agents, they remain constrained by the standard chat interface, which lacks private working memory. This raises a fundamental question: can agents reliably perform interactive tasks that depend on hidden state? We define Private State Interactive Tasks (PSITs), which require agents to generate and maintain hidden information while producing public responses consistent with a fixed hidden state. We show theoretically that any agent restricted to the public conversation history cannot both keep the secret unresolved in the transcript and respond consistently with a fixed hidden state in PSITs, yielding an architectural impossibility theorem. …

4. Close Builds on this paper Not relevant

   A Survey on Large Language Models for Automated Planning

   Mohamed Aghzal, E. Plaku, Gregory J. Stein et al. · 2025 · arXiv.org · preprint, not peer reviewed · cited 31 times · Not summarized yet

   A Survey on Large Language Models for Automated Planning cites this paper in its review of LLM planning successes and shortcomings.

   Abstract

   The planning ability of Large Language Models (LLMs) has garnered increasing attention in recent years due to their remarkable capacity for multi-step reasoning and their ability to generalize across a wide range of domains. While some researchers emphasize the potential of LLMs to perform complex planning tasks, others highlight significant limitations in their performance, particularly when these models are tasked with handling the intricacies of long-horizon reasoning. In this survey, we critically investigate existing research on the use of LLMs in automated planning, examining both their successes and shortcomings in detail. We illustrate that although LLMs are not well-suited to serve …

5. Close Broader context Not relevant

   Inducing state anxiety in LLM agents reproduces human-like biases in consumer decision-making

   Ziv Ben-Zion, Z. Elyoseph, Tobias R. Spiller et al. · 2025 · npj Artificial Intelligence · cited 6 times · Not summarized yet

   Inducing state anxiety studies how emotional context changes agents’ consumer actions, a different test of factors affecting agent behavior.

   Abstract

   Large language models (LLMs) are rapidly evolving from text generators to autonomous agents, raising urgent questions about their reliability in real-world contexts. A central question is whether emotionally salient context can systematically steer LLM agents’ action policies, not only their text outputs, in applied tasks. Here, three advanced LLMs (ChatGPT-5, Gemini 2.5, Claude 3.5-Sonnet) performed a grocery shopping task under budget constraints, before and after exposure to anxiety-inducing traumatic narratives. Across 2,250 runs, emotionally primed agents consistently selected less healthy baskets (as quantified by decreases in Basket Health Scores: Δ=-0.081 to -0.126; all pFDR < 0.001;…