arXiv:2606.04484v1 Announce Type: cross 
Abstract: We present AgentJet, a distributed swarm training framework for large language model (LLM) agent reinforcement learning. Unlike centralized frameworks that tightly couple agent rollouts with model optimization, AgentJet adopts a decoupled multi-node architecture in which swarm server nodes host trainable models and run optimization on GPU clusters, whereas swarm client nodes execute arbitrary agents on arbitrary devices. This design provides capabilities that are difficult to support in centralized frameworks: (1) heterogeneous multi-model reinforcement learning, enabling the training of heterogeneous multi-agent teams with multiple LLM as brains; (2) multi-task cocktail training with isolated agent runtimes; (3) fault-tolerant execution that prevents external environment failures from interrupting the training process; and (4) live code iteration, which allows agents to be edited during training by replacing swarm client nodes. To support efficient RL in multi-model, multi-turn, and multi-agent settings, AgentJet introduces a context tracking module with timeline merging, which consolidates redundant context and achieves a 1.5-10x training speedup. Finally, AgentJet introduces an automated research system that takes a research topic as input and autonomously conducts long-horizon, multi-day RL studies on large-scale clusters. By leveraging the swarm architecture, this system reproduces key exploratory workflows of RL researchers without human intervention during execution.

تم تقديم AgentJet كإطار تدريب جماعي موزع مصمم لتعلم التعزيز لوكلاء نماذج اللغة الكبيرة (LLM)، مع بنية مفصولة تفصل بين تحسين النموذج وتنفيذ الوكيل. يسمح هذا الإطار بتدريب متعدد الوكلاء غير متجانس، وتحمل الأخطاء، وتكرار التعليمات البرمجية الحية، مما يعزز قدرات LLM في بيئات متنوعة.

AgentJet se ha presentado como un marco de entrenamiento en enjambre distribuido diseñado para el aprendizaje por refuerzo de agentes de modelos de lenguaje de gran tamaño (LLM), con una arquitectura desacoplada que separa la optimización del modelo de la ejecución del agente. Este marco permite el entrenamiento heterogéneo de múltiples agentes, la tolerancia a fallos y la iteración de código en vivo, mejorando así las capacidades de los LLM en diversos entornos.

AgentJet a été introduit comme un cadre d'entraînement en essaim distribué conçu pour l'apprentissage par renforcement des agents de modèles de langage de grande taille (LLM), présentant une architecture découplée qui sépare l'optimisation des modèles de l'exécution des agents. Ce cadre permet un entraînement hétérogène multi-agents, une tolérance aux pannes et une itération de code en direct, améliorant ainsi les capacités des LLM dans divers environnements.

AgentJet has been introduced as a distributed swarm training framework designed for large language model (LLM) agent reinforcement learning, featuring a decoupled architecture that separates model optimization from agent execution. This framework allows for heterogeneous multi-agent training, fault tolerance, and live code iteration, enhancing the capabilities of LLMs in various environments.