arXiv AI

CollabFlow: Recursive Self-Improvement of Agent Collaboration

CollabFlow introduces a recursive self‑improvement framework for multi‑agent collaboration in large language model systems. It trains a Collab‑Director to assemble teams of agents, uses a frozen executor to run them, and retrains the director each round based on outcomes. The system incorporates evidence‑conditioned communication protocols within collaboration graphs and a Collaborative Trajectory Balance objective to maintain diverse high‑performing teams across rounds, achieving superior performance on twelve datasets.

arXiv AI
Oct 1

EvoSteer: Online Self-Evolving Graph Orchestration via Reference-Anchored Credit Assignment

EvoSteer introduces an online self‑evolving graph orchestration framework that continuously builds and repairs a team of tool‑using agents during execution. It employs Anchored Trajectory Balance (AnchorTB), a regression‑style loss that assigns credit to each orchestration action by comparing subtrajectories to a frozen reference, and Validated Skill Admission, which tests candidate skills before promotion. Experiments on twelve datasets demonstrate that EvoSteer outperforms existing baselines in question answering, mathematical reasoning, code generation, and interactive decision making.

By Mingda Zhang, Hanwen Zhang, Qiang Huang, Zijia Wang, Pengfei Guo, Yuchen Zhang, Jionghao Zhu, Xiaoying Tang
arXiv Computation and Language
Sep 23

CONCAT: Consensus- and Confidence-Driven Ad Hoc Teaming for Efficient LLM-Based Multi-Agent Systems

CONCAT is a training‑free framework that improves the efficiency of large language model (LLM) based multi‑agent systems by clustering agents according to their initial answers and selecting cluster leaders based on confidence. It uses a Theory‑of‑Mind‑inspired heuristic to predict collaboration benefits between leaders, then prunes communications to form an ad‑hoc network that reduces latency. Experiments on three LLMs and benchmarks show up to 2.02× higher accuracy/latency ratio than LLM‑Debate and a 50.1% latency reduction on Qwen2.5‑14B‑Instruct without task‑specific training.

By Ziyang Ma, Dingyi Zhang, Sichu Liang, Jiajia Chu, Pengfei Xia, Hui Zang, Deyu Zhou
arXiv AI
Oct 15

CoEvoke: Evoking Multi-Agent Collaboration for Trajectory Collection

arXiv:2610.14021v1 Announce Type: cross Abstract: Training multi-agent collaboration requires trajectories of checking, feedback, and repair, yet task-solving rollouts may end in immediate approval o...

By Lin Wu, Shu Yao, Yuheng Wang, Jingru Fan, Huatao Li, Yihan Li, Liu Yixuan, Cheng Liu, Jian Chang, Zeyu Zhao, Aohua Zhang, Huiying Zhao, Shuyuan Zhu, Chen Qian
arXiv Computation and Language
Oct 16

NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness

NeoHorse-1 is a family of agent-native models that pursue recursive self‑improvement through agentic post‑training. The system uses a heterogeneous model pool with intelligent routing to record capability demands, service tiers, and interactions, converting these records into training examples that preserve reasoning, tool calls, and harness context. By structuring supervised fine‑tuning into a three‑stage curriculum and employing routing‑guided on‑policy distillation, NeoHorse-1 closes an evaluation‑selection‑update loop, raising macro‑average performance from 58.94 to 64.87 on 4B models and from 65.60 to 69.04 on 9B models across eleven benchmarks, while also producing a prefill‑only variant that outputs structured decisions and probabilities for agent workflows.

By NeoHorse Team, Guoliang Cao, Guohao Dai, Tianyu Guo, Kai Han, Hailin Hu, Zihan Jiang, Xiang Kuang, Boxun Li, Yulong Li, Zehua Pei, Yuchuan Tian, Jiamin Wang, Yu Wang, Yunhe Wang, Yihong Wu, Haiyang Xu, Shuo Zhang, Hang Zhou, Siyang Cheng, Jiayu Fan, Wei He, Qingrui Jiao, Hongguang Li, Zhiyuan Li, Runke Liu, Xi Liu, Xinchen Liu, Sinno Jialin Pan, Yi Ren, Liuyang Song, Chenyu Wang, Bei Yu, Quanlu Zhang, Xiangyu Zhang, Mengyu Zheng, Yingjie Zong
arXiv AI
Jul 23

In-the-Flow Agentic System Optimization for Effective Planning and Tool Use

arXiv:2510. 05592v2 Announce Type: replace Abstract: Outcome-driven reinforcement learning has advanced reasoning in large language models (LLMs), but prevailing tool-augmented approaches train a single, monolithic policy that interleaves thoughts and tool calls under full context; this scales poorly with long horizons and diverse tools and generalizes weakly to new scenarios.

By Zhuofeng Li, Haoxiang Zhang, Seungju Han, Sheng Liu, Jianwen Xie, Yu Zhang, Yejin Choi, James Zou, Pan Lu
Hugging Face Trending Papers
Jun 18

Connect the Dots: Training LLMs for Long-Lifecycle Agents with Cross-Domain Generalization Via Reinforcement Learning

This work presents a general framework for training large language models (LLMs) to "Connect the Dots" (CoD), a meta-capability required by long-lifecycle agents: as an LLM-based AI agent gets deployed in an environment, it solves a long sequence of tasks while continuously exploring the environment, learning from its own experiences, and iteratively self-updating its context about the environment, thereby achieving progressively better performance on future tasks conditioned on the updated context. Major components of the CoD framework include: (1) algorithm design and infrastructure for end-to-end reinforcement learning (RL) with long rollout sequences interleaving solve-task and update-context episodes; (2) tasks and environments for incentivizing and eliciting the targeted meta-capability in LLMs during training, as well as for faithfully measuring progress during evaluation.

arXiv AI
Oct 31

MAS-Loop: Beyond One-Shot Orchestration towards Self-Evolving Multi-Agent Systems

MAS-Loop introduces an execution‑grounded framework that couples an Orchestrator with a trajectory‑conditioned Refiner to close the feedback loop in training‑time multi‑agent system design. For each task, the Orchestrator builds and runs a source MAS, then the Refiner uses the resulting trajectory to locate failures, diagnose them, and propose bounded patches to agents or execution settings. The updated MAS is re‑executed, and the pairwise comparison of outcomes provides learning signals that enable both components to co‑evolve, yielding stronger systems and a new dataset, MAS‑JudgeCorpus, for trajectory‑aware MAS research.

By Hehai Lin, Qi Yang, Chengwei Qin
arXiv AI
Jul 29

Toward an Organizational Science of Multi-Agent LLM Systems: Decoupling Who, How, and Which Algorithm

arXiv:2607. 25446v1 Announce Type: new Abstract: Multi-agent frameworks built on large language models (LLMs) routinely entangle three logically distinct concerns: who is on the team (organization), how members align (coordination), and which algorithm fuses their work (collaboration protocol).

By Huan Chen, Xiang Song, Jian Jin, Pan Ren, Liang-Jie Zhang
arXiv AI
3d ago

From Delegation to Coordination: Learning Efficient Multi-Agent Deep Research

The paper introduces LEAD-RL, a joint training framework for multi‑agent deep research systems that assigns credit to subagents based on their contribution to the main agent’s research utility. By evaluating each subagent’s report during training and gating the inherited team advantage, LEAD‑RL avoids rewarding misleading subagent outputs and focuses on final‑answer correctness. Experiments on five benchmarks and two model families show that LEAD‑RL improves answer quality, achieving a 2.49‑point accuracy gain on BrowseComp and a 24.5% reduction in critical‑path cost compared to shared‑credit joint training.

By Yuyang Wu, Yang Liu, Zequn Xie, Junjie Wang, Dan Yang, Zehao Tan, Ding-Chu Zhang, Junming Yang, Yue Shen, Jian Wang, Jinjie Gu
arXiv AI
Sep 11

ROTATE: Regret-driven Open-ended Training for Ad Hoc Teamwork

The paper introduces ROTATE, a regret-driven open‑ended training framework that jointly improves an Ad Hoc Teamwork (AHT) agent and an adversarial teammate generator. Unlike traditional two‑stage pipelines, ROTATE alternates between enhancing the agent and generating teammates that specifically probe its collaboration weaknesses. Experiments on Overcooked and Level‑Based Foraging show that ROTATE outperforms existing baselines on unseen teammates, setting a new benchmark for robust, generalizable teamwork.

By Caroline Wang, Arrasy Rahman, Benjamin Nativi, Jiaxun Cui, Yoonchang Sung, Peter Stone
arXiv AI
Jun 19

Connect the Dots: Training LLMs for Long-Lifecycle Agents with Cross-Domain Generalization Via Reinforcement Learning

arXiv:2606. 20002v1 Announce Type: cross Abstract: This work presents a general framework for training large language models (LLMs) to "Connect the Dots" (CoD), a meta-capability required by long-lifecycle agents: as an LLM-based AI agent gets deployed in an environment, it solves a long sequence of tasks while continuously exploring the environment, learning from its own experiences, and iteratively self-updating its context about the environment, thereby achieving progressively better performance on future tasks conditioned on the updated context.

By Yanxi Chen, Weijie Shi, Yuexiang Xie, Boyi Hu, Yaliang Li, Bolin Ding, Jingren Zhou