AgentJet: Swarm framework trains LLM agents 10x faster

AgentJet is a new distributed swarm training framework for LLM-based agent reinforcement learning, described in a technical report on arXiv. Unlike centralized frameworks that tightly couple agent rollouts with model optimization, AgentJet adopts a decoupled multi-node architecture. Swarm server nodes host trainable models and run optimization on GPU clusters, while swarm client nodes execute arbitrary agents on arbitrary devices. This design enables heterogeneous multi-model reinforcement learning—training teams of agents powered by different LLMs—and multi-task cocktail training with isolated runtimes. It also provides fault-tolerant execution, preventing external environment failures from interrupting training, and live code iteration, allowing agents to be edited on the fly by replacing client nodes.

To handle the complexity of multi-model, multi-turn, multi-agent settings, AgentJet introduces a context tracking module with timeline merging that consolidates redundant context, achieving a 1.5-10x training speedup. The framework also includes an automated research system that takes a research topic as input and autonomously conducts long-horizon, multi-day RL studies on large-scale clusters, reproducing key exploratory workflows of RL researchers without human intervention. This work was submitted by Qingxu Fu, Boyin Liu, Shuchang Tao, Zhaoyang Liu, and Bolin Ding.