Cutting-edge agentic AI systems are built on foundation models that can be adapted to plan, reason, and interact with external tools to perform increasingly complex and specialized tasks. As these systems grow in capability and scope, adaptation becomes a central mechanism for improving performance, reliability, and generalization. In this paper, we unify the rapidly expanding research landscape into a systematic framework that spans both agent adaptations and tool adaptations. We further decompose these into tool-execution-signaled and agent-output-signaled forms of agent adaptation, as well as agent-agnostic and agent-supervised forms of tool adaptation. We demonstrate that this framework helps clarify the design space of adaptation strategies in agentic AI, makes their trade-offs explicit, and provides practical guidance for selecting or switching among strategies during system design. We then review the representative approaches in each category, analyze their strengths and limitations, and highlight key open challenges and future opportunities. Overall, this paper aims to offer a conceptual foundation and practical roadmap for researchers and practitioners seeking to build more capable, efficient, and reliable agentic AI systems.
Figure 2 The structure of this paper.
The rapid progress of foundation models, such as large language models (LLMs), has catalyzed the rise of agentic AI systems: autonomous AI systems capable of perceiving their environment, invoking external tools, managing memory, and executing multi-step plans toward completing complex tasks [1][2][3][4]. Agentic AI demonstrates remarkable potential in applications ranging from scientific discovery [5,6] to software development and clinical research [7][8][9]. However, current agentic AI systems still struggle with challenges such as unreliable tool use, limited long-horizon planning, domain-specific reasoning gaps, robustness issues in real-world environments, and poor generalization to unexplored environments where the agent lacks prior interaction experience [10][11][12][13]. These limitations reveal that even highly capable foundation models often require additional adaptation to specialize for particular tasks or real-world scenarios. This motivates the need for adaptation in agentic AI systems, whereby the components of an agentic system are modified or optimized so that the agent achieves higher task performance, improved reliability, and better generalization across diverse scenarios.
Building on this motivation, we conduct a comprehensive survey on the adaptation in agentic AI systems, aiming to systematically analyze how components in agentic AI systems are modified to overcome current limitations. Compared with existing surveys on modern AI agents [1,[14][15][16][17][18], this paper centers specifically on adaptation in agentic AI. To structure this rapidly expanding literature, we introduce a unified framework that organizes adaptation in agentic AI into four core paradigms spanning both agent adaptation and tool adaptation, as shown in Figure 1. This framework clarifies the underlying design space, highlights the trade-offs between different adaptation strategies, and provides practical guidance for choosing or transitioning between paradigms based on supervision signals, task requirements, and system-level constraints.
In our framework, we conclude adaptation strategies for agentic AI into two dimensions according to which component is optimized ( §3). The first dimension, which we term Agent Adaptation, focuses on modifying the agent’s internal parameters, representations, or behavioral policies to better align with task requirements. This includes both traditional fine-tuning approaches [19] and modern reinforcement learning methods that leverage environment feedback [20,21]. The second dimension, Tool Adaptation, shifts the optimization target from the agent to its external tools, e.g., retrievers, planners, memory modules, and specialized models, enabling frozen agents to benefit from an adaptive operational environment [22,11,23]. Within these two broad paradigms, we further identify four distinct adaptation strategies, forming a comprehensive taxonomy that organizes the rapidly evolving landscape of agentic AI research:
• A1: Tool Execution Signaled Agent Adaptation ( §3.2.1, §4.1): The agent is optimized using verifiable outcomes produced by external tools it invokes. This paradigm captures settings where correctness signals arise directly from tool execution, such as code sandbox results, retrieval relevance scores, or API call outcomes.
• A2: Agent Output Signaled Agent Adaptation ( §3.2.2, §4.2): The agent is optimized using evaluations of its own outputs, e.g., final answers, plans, or reasoning traces, possibly after incorporating tool results. This paradigm includes both tool-free outcome-based learning and tool-augmented adaptation driven by answer correctness or preference scores.
• T1: Agent-Agnostic Tool Adaptation ( §3.2.3, §5.1): Tools are trained independently of the frozen agent. These tools include retrievers, domain-specific models, and other pretrained components that can be used as plug-and-play modules orchestrated by the frozen agent.
• T2: Agent-Supervised Tool Adaptation ( §3.2.4, §5.2): The agent remains fixed while its tools are adapted using signals derived from the agent’s outputs. This paradigm includes reward-driven retriever tuning, adaptive rerankers, search subagents, and memory-update modules trained to better support the frozen agent.
It is worth noting that these four strategies are not mutually exclusive: state-of-the-art systems increasingly combine multiple adaptation paradigms to achieve optimal performance [24][25][26]. For instance, a deep research system might employ T1-style retrieval tools (pre-trained dense retrievers), T2-style adaptive search agents (trained via frozen LLM feedback), and A1-style reasoning agents (fine-tuned with execution feedback) in a cascaded architecture [6].
In §6, we further emphasize that the choice among these paradigms involves fundamental trade-offs along several dimensions. (1) Cost and flexibility: Agent adaptation (A1/A2) typically requires substantial computationa
This content is AI-processed based on open access ArXiv data.
