Agentic Reasoning for Large Language Models
by arXiv (Tianxin Wei et al.)
A 2026 survey mapping how reasoning drives planning, tool use, memory, and multi-agent coordination in LLM agents.
Overview
Submitted to arXiv on January 18, 2026 (arXiv:2601.12538) by a 29-author group led by Tianxin Wei, this survey examines how large language models act as autonomous agents through interaction with their environment. It structures the field into foundational agentic reasoning (planning and tool use in stable settings), self-evolving reasoning (memory and adaptation via feedback), and multi-agent reasoning (coordination and knowledge sharing), and it distinguishes in-context reasoning methods from post-training optimization through reinforcement learning. The paper surveys applications across science, robotics, healthcare, autonomous research, and mathematics, and closes by naming open challenges including personalization, long-horizon tasks, world modeling, and deployment governance. It is a strong orientation and reference for anyone building or researching reasoning-driven agents.
At a Glance
- Topic
- Agentic
- Level
- Advanced
- Format
- Paper
- Cost
- Free
- Duration
- Survey paper (long read)
- Provider
- arXiv (Tianxin Wei et al.)
- Hands-on
- No
- Certificate
- None
What You’ll Learn
- ✓A taxonomy of agentic reasoning: foundational, self-evolving, and multi-agent
- ✓How planning, tool use, memory, and feedback combine in LLM agents
- ✓The difference between in-context reasoning and RL-based post-training for agents
- ✓Where agentic reasoning is being applied across science, robotics, and research
- ✓Open problems: long-horizon tasks, world modeling, personalization, and governance
Highlights
- •Recent (Jan 2026) and broad, covering both methods and applications
- •Clear three-part taxonomy that maps a sprawling literature
- •Rich reference list for going deeper on any subtopic
Who It’s For
Best For
- ✓Researchers surveying the agentic-reasoning literature
- ✓AI engineers wanting a conceptual map before building agents
- ✓Graduate students entering LLM-agent research
Prerequisites
- •Comfort reading ML research papers
- •Working knowledge of LLMs, agents, and reinforcement learning basics
FAQ
What is Agentic Reasoning for Large Language Models?
A January 2026 arXiv survey that organizes the fast-moving field of agentic reasoning in large language models. It is for researchers and AI engineers who want a structured map of how planning, tool use, memory, feedback-driven adaptation, and multi-agent coordination fit together.
Is Agentic Reasoning for Large Language Models free?
Agentic Reasoning for Large Language Models is free to access.
What level is Agentic Reasoning for Large Language Models for?
Agentic Reasoning for Large Language Models is aimed at a advanced audience. Recommended background: Comfort reading ML research papers, Working knowledge of LLMs, agents, and reinforcement learning basics.
How long does Agentic Reasoning for Large Language Models take?
Expect roughly Survey paper (long read). Most learners work through it at their own pace.
What will I learn from Agentic Reasoning for Large Language Models?
You'll learn: A taxonomy of agentic reasoning: foundational, self-evolving, and multi-agent; How planning, tool use, memory, and feedback combine in LLM agents; The difference between in-context reasoning and RL-based post-training for agents; Where agentic reasoning is being applied across science, robotics, and research; Open problems: long-horizon tasks, world modeling, personalization, and governance.