The Year in AI Papers
The year/Topics/Agents

Topic area

Agents

Every collection across agents.

Papers
275
Research labs
10
Official code
214

251–275 of 275 papers in this topic area

251

Qwen

Scaling Agents via Continual Pre-training

The paper introduces Agentic Continual Pre-training (Agentic CPT), a new training stage between pre-training and post-training, to build agentic foundation models for deep research agents. The authors argue that post-training on general-purpose models creates optimization conflicts, as models must simultaneously learn agentic behaviors and align to expert…

Liangcai Su, Zhen Zhang, Guangyu Li, Zhuo Chen, et al.
Published
Sep 2025
Citations
41
Code
20K stars
Read summary Original
252

arXiv.org

WebResearcher: Unleashing unbounded reasoning capability in Long-Horizon Agents

WebResearcher is a framework for deep-research agents that addresses the limitations of mono-contextual approaches, which suffer from context suffocation and noise contamination. It introduces IterResearch, an iterative paradigm that reformulates deep research as a Markov Decision Process, where agents periodically consolidate findings into evolving…

Zile Qiao, Guoxin Chen, Xuanzhong Chen, Donglei Yu, et al.
Published
Sep 2025
Citations
66
Code
Not linked
Read summary Original
253

arXiv.org

WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learning

WebSailor-V2 is a post-training pipeline for open-source web agents, comprising data construction (SailorFog-QA-V2), SFT, and RL. The dataset uses a dense knowledge graph with cyclic structures and diverse uncertainty types beyond obfuscation. Training employs a dual-environment RL framework: a high-fidelity simulator for rapid iteration and a managed…

Kuan Li, Zhongwang Zhang, Huifeng Yin, Rui Ye, et al.
Published
Sep 2025
Citations
60
Code
Not linked
Read summary Original
254

arXiv.org

AgentGym-RL: Training LLM Agents for Long-Horizon Decision Making through Multi-Turn Reinforcement Learning

AgentGym-RL is a unified, modular reinforcement learning framework for training LLM agents in multi-turn, long-horizon decision-making tasks, without requiring supervised fine-tuning. It supports diverse environments (web navigation, deep search, digital games, embodied tasks, scientific tasks) and mainstream RL algorithms (PPO, GRPO, REINFORCE++, RLOO).…

Zhiheng Xi, Jixuan Huang, Chenyang Liao, Baodai Huang, et al.
Published
Sep 2025
Citations
62
Code
834 stars
Read summary Original
255

arXiv.org

Mini-o3: Scaling Up Reasoning Patterns and Interaction Turns for Visual Search

Mini-o3 is a system for visual search that scales up tool-based interactions and reasoning patterns, achieving state-of-the-art performance on challenging tasks. It addresses limitations of existing open-source VLMs, which show monotonous reasoning and limited interaction turns. The approach includes three key components: constructing the Visual Probe…

Xin Lai, Junyi Li, Wei Li, Tao Liu, et al.
Published
Sep 2025
Citations
101
Code
424 stars
Read summary Original
256

arXiv.org

WebExplorer: Explore and Evolve for Training Long-Horizon Web Agents

WebExplorer introduces a data synthesis framework for training long-horizon web agents. It uses model-based exploration, where LLMs iteratively search and browse from a seed entity to construct an information space, and iterative long-to-short query evolution, which removes salient clues and adds obfuscation to increase query difficulty. This produces the…

Junteng Liu, Yunji Li, Chi Zhang, Jingyang Li, et al.
Published
Sep 2025
Citations
74
Code
120 stars
Read summary Original
257

arXiv.org

Inverse IFEval: Can LLMs Unlearn Stubborn Training Conventions to Follow Real Instructions?

The paper introduces Inverse IFEval, a benchmark to evaluate LLMs' ability to follow counterintuitive instructions that conflict with training conventions. It identifies 'cognitive inertia' from SFT, where models struggle with instructions deviating from standard patterns. The benchmark includes eight instruction types (e.g., Question Correction,…

Qinyan Zhang, Xinping Lei, Ruijie Miao, Yu Fu, et al.
Published
Sep 2025
Citations
6
Code
Not linked
Read summary Original
258

arXiv.org

UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning

UI-TARS-2 is a native GUI-centered agent model developed by ByteDance Seed, designed to handle both structured computer-use tasks and dynamic game environments. It addresses challenges in data scalability, multi-turn reinforcement learning (RL), GUI-only operation limits, and environment stability through a systematic methodology: a data flywheel for…

Haoming Wang, Haoyang Zou, Huatong Song, Jiazhan Feng, et al.
Published
Sep 2025
Citations
163
Code
11K stars
Read summary Original
259

AAAI Conference on Artificial Intelligence

DeepResearch Arena: The First Exam of LLMs' Research Abilities via Seminar-Grounded Tasks

The paper introduces DeepResearch Arena, a benchmark for evaluating deep research agents using tasks derived from academic seminar transcripts. It addresses limitations of existing benchmarks, which either risk data leakage from static corpora or lack scalability and realism in expert-curated sets. The authors propose a Multi-Agent Hierarchical Task…

Haiyuan Wan, Chen Yang, Junchi Yu, Meiqi Tu, et al.
Published
Sep 2025
Citations
26
Code
Not linked
Read summary Original
260

arXiv.org

VerlTool: Towards Holistic Agentic Reinforcement Learning with Tool Use

VERLTOOL is a unified, modular framework for Agentic Reinforcement Learning with Tool use (ARLT), addressing fragmentation, synchronous execution bottlenecks, and limited extensibility in existing ARLT codebases. It builds on VeRL, providing upstream alignment, a standardized tool server API supporting diverse tools (code execution, search, SQL, vision,…

Dongfu Jiang, Yi Lu, Zhuofeng Li, Zhiheng Lyu, et al.
Published
Sep 2025
Citations
81
Code
1K stars
Read summary Original
261

arXiv.org

Open Data Synthesis For Deep Research

This technical report introduces InfoSeek, a scalable framework for synthesizing Deep Research tasks, formalized as Hierarchical Constraint Satisfaction Problems (HCSPs). Unlike simpler multi-hop or flat CSP problems, HCSPs require traversing a hierarchy of interdependent constraints. InfoSeek uses a dual-agent system (Planner and Browser) to recursively…

Ziyi Xia, Kun Luo, Hongjin Qian, Zheng Liu
Published
Aug 2025
Citations
14
Code
216 stars
Read summary Original
262

arXiv.org

MCP-Bench: Benchmarking Tool-Using LLM Agents with Complex Real-World Tasks via MCP Servers

MCP-Bench is a benchmark for evaluating LLM agents on realistic, multi-step tool-use tasks via the Model Context Protocol (MCP). It connects agents to 28 live MCP servers with 250 tools across domains like finance, travel, and science, enabling complex intra- and cross-server workflows. Tasks are generated via an LLM-based synthesis pipeline that discovers…

Zhenting Wang, Qi Chang, Hemani Patel, Shashank Biju, et al.
Published
Aug 2025
Citations
90
Code
496 stars
Read summary Original
263

Annual Meeting of the Association for Computational Linguistics

A.S.E: A Repository-Level Benchmark for Evaluating Security in AI-Generated Code

The paper introduces A.S.E (AI Code Generation Security Evaluation), a repository-level benchmark for assessing the security of AI-generated code. It is built from 120 instances derived from 40 real-world GitHub repositories with documented CVEs, expanded via semantic and structural mutations. The benchmark covers four vulnerability types (SQL injection,…

Keke Lian, Bin Wang, Lei Zhang, Libo Chen, et al.
Published
Aug 2025
Citations
15
Code
647 stars
Read summary Original
264

arXiv.org

AgentScope 1.0: A Developer-Centric Framework for Building Agentic Applications

AgentScope 1.0 is a developer-centric framework for building agentic applications, grounded in the ReAct paradigm. It provides foundational components (message, model, memory, tool) with unified interfaces and extensible modules, supporting multimodal messages, diverse LLM providers, and MCP integration. The framework includes agent-level infrastructure…

Dawei Gao, Zitao Li, Yuexiang Xie, Weirui Kuang, et al.
Published
Aug 2025
Citations
11
Code
29K stars
Read summary Original
265

arXiv.org

Memento: Fine-tuning LLM Agents without Fine-tuning LLMs

The paper introduces Memento, a learning paradigm for LLM agents that enables continual adaptation without fine-tuning the underlying LLM. It formalizes a Memory-augmented Markov Decision Process (M-MDP) with a neural case-selection policy, storing past experiences in an episodic case bank. The policy is updated via online soft Q-learning, with retrieval…

Huichi Zhou, Yihang Chen, Siyuan Guo, Xue Yan, et al.
Published
Aug 2025
Citations
87
Code
2.6K stars
Read summary Original
266

arXiv.org

Mobile-Agent-v3: Fundamental Agents for GUI Automation

The paper introduces GUI-Owl, a foundational GUI agent model built on Qwen2.5-VL, and Mobile-Agent-v3, a multi-agent framework. GUI-Owl-7B achieves state-of-the-art scores among open-source models, including 66.4 on AndroidWorld and 34.9 on OSWorld-Verified, while Mobile-Agent-v3 further improves these to 73.3 and 37.7, respectively. Key innovations…

Jiabo Ye, Xi Zhang, Haiyang Xu, Haowei Liu, et al.
Published
Aug 2025
Citations
150
Code
9K stars
Read summary Original
267

arXiv.org

FutureX: An Advanced Live Benchmark for LLM Agents in Future Prediction

FutureX is a live benchmark for evaluating LLM agents on future prediction tasks, built to avoid data contamination by using only future events. It collects questions daily from 195 curated websites across 11 domains, runs 25 models (base LLMs, Think&Search models, open-source and closed-source Deep Research agents), and automatically scores predictions…

Zhiyuan Zeng, Jiashuo Liu, Siyuan Chen, Tianci He, et al.
Published
Aug 2025
Citations
39
Code
Not linked
Read summary Original
268

AAAI Conference on Artificial Intelligence

ComoRAG: A Cognitive-Inspired Memory-Organized RAG for Stateful Long Narrative Reasoning

ComoRAG is a cognitive-inspired, memory-organized RAG framework for stateful long narrative reasoning. It addresses the limitations of traditional RAG methods, which are stateless and single-step, by mimicking the human Prefrontal Cortex's Metacognitive Regulation. The framework uses a dynamic memory workspace and an iterative cognitive loop with five…

Juyuan Wang, Rongchen Zhao, Wei Wei, Yufeng Wang, et al.
Published
Aug 2025
Citations
13
Code
343 stars
Read summary Original
269

arXiv.org

WideSearch: Benchmarking Agentic Broad Info-Seeking

The paper introduces WideSearch, a benchmark for evaluating LLM-based search agents on large-scale, broad information-seeking tasks. It comprises 200 manually curated questions (100 English, 100 Chinese) across 18 domains, requiring agents to collect and structure extensive atomic information into tables. A five-stage quality control pipeline ensures task…

Ryan Wong, Jiawei Wang, Junjie Zhao, Li Chen, et al.
Published
Aug 2025
Citations
53
Code
149 stars
Read summary Original
270

arXiv.org

A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems

This survey reviews the emerging paradigm of self-evolving AI agents, which aim to bridge static foundation models with lifelong adaptability. It introduces a unified framework with four components: system inputs, agent system, environment, and optimizers, and proposes 'Three Laws' (Endure, Excel, Evolve) to guide safe and effective evolution. The survey…

Jinyuan Fang, Yanwen Peng, Xi Zhang, Yingxu Wang, et al.
Published
Aug 2025
Citations
154
Code
2.4K stars
Read summary Original
271

Qwen

WebWatcher: Breaking New Frontier of Vision-Language Deep Research Agent

WebWatcher is a multimodal deep research agent introduced by Alibaba's Tongyi Lab to address the limitations of text-centric web agents by integrating visual information and complex reasoning. It uses high-quality synthetic multimodal trajectories for cold-start training, multiple tools (web search, image search, webpage visit, code interpreter, OCR), and…

Xinyu Geng, Peng Xia, Zhen Zhang, Xinyu Wang, et al.
Published
Aug 2025
Citations
101
Code
20K stars
Read summary Original
272

arXiv.org

Chain-of-Agents: End-to-End Agent Foundation Models via Multi-Agent Distillation and Agentic RL

The paper introduces Chain-of-Agents (CoA), a paradigm enabling a single LLM to perform multi-agent problem-solving end-to-end by dynamically activating role-playing and tool agents. To train these Agent Foundation Models (AFMs), the authors propose multi-agent distillation, which converts trajectories from state-of-the-art multi-agent systems (e.g.,…

Weizhen Li, Jianbo Lin, Zhuosong Jiang, Jingyi Cao, et al.
Published
Aug 2025
Citations
66
Code
580 stars
Read summary Original
273

Independent research

VeriWeb: Verifiable Long-Chain Web Benchmark for Agentic Information-Seeking

VeriWeb is a new benchmark for evaluating web agents on long-horizon, information-rich tasks. It addresses limitations of prior benchmarks that focus on single-fact retrieval and outcome-only verification. VeriWeb emphasizes long-chain complexity (breadth- and depth-oriented search) and subtask-level verifiability, decomposing tasks into interdependent…

Shunyu Liu, Minghao Liu, Huichi Zhou, Zhenyu Cui, et al.
Published
Aug 2025
Citations
9
Code
88 stars
Read summary Original
274

arXiv.org

Agent Lightning: Train ANY AI Agents with Reinforcement Learning

Agent Lightning is a framework that enables reinforcement learning (RL) training of large language models (LLMs) for any AI agent, achieving complete decoupling between agent execution and training with almost zero code modifications. It formulates agent execution as a Markov decision process (MDP), defining a unified data interface that abstracts agent…

Xufang Luo, Yuge Zhang, Zhiyuan He, Zilong Wang, et al.
Published
Aug 2025
Citations
58
Code
17K stars
Read summary Original
275

arXiv.org

Training Long-Context, Multi-Turn Software Engineering Agents with Reinforcement Learning

The paper presents a two-phase training pipeline for software engineering (SWE) agents using reinforcement learning (RL). Starting from Qwen2.5-72B-Instruct, the authors first apply rejection fine-tuning (RFT) on successful trajectories from SWE-rebench tasks, improving Pass@1 on SWE-bench Verified from 11% to 20%. Then, they apply a synchronous RL…

Alexander Golubev, Maria Trofimova, Sergei Polezhaev, Ibragim Badertdinov, et al.
Published
Aug 2025
Citations
27
Code
Not linked
Read summary Original