Prime Agent: A Self-Improving RLM Harness
Prime Agent:一种自改进的RLM管控框架
机构 * Princeton University(普林斯顿大学) ; MIT(麻省理工学院) ; Prime Intellect
AI总结 Prime Agent是一款开源RLM管控框架,通过标准化流程提升模型长周期能力,在ARC-AGI-3等任务中大幅提升性能,支持编码等工作流与并行化任务。
Comments 16 pages, 10 figures. Technical report. Code: this https URL (https://github.com/PrimeIntellect-ai/prime-agent)