发表机构
NVIDIA; NTU; MIT（英伟达; 南洋理工大学; 麻省理工学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
SoL-Pi通过递归扩展自动研究循环，在框架层提升令牌效率，在EdgeBench上以44.7-49.0%的令牌减少和约三分之一成本降低实现与Pi相当的性能。
AI 中文摘要
随着编码智能体从有监督的代码补全转向无人值守的全天候探索，其工作范围从孤立的预测扩展为包含推理、工具使用和反馈的长轨迹。因此，令牌效率对于扩展递归自我改进变得至关重要。我们在框架层采用受RSI启发的方法，在越来越多的多样化环境中扩展自动研究循环以进行框架部署。在此规模下，该过程产生可复用的改进，这些改进能超越其开发环境进行迁移，推动自动化框架发现走向生产级成果。四种机制在筛选中幸存下来并构成SoL-Pi，涵盖动作执行、上下文压缩、观察处理和委派阅读。在51项任务的EdgeBench评估中，SoL-Pi在GPT-5.6 Sol和Opus 5上实现了与Pi相当的性能，同时将记录的令牌流量减少了44.7-49.0%，API成本降低了约三分之一。换言之，相对于原生Codex和Claude Code框架，估计每小时节省8.75-13.50美元，相对于Pi节省4.36-5.71美元。
英文摘要
As coding agents move from supervised code completion to unattended, around-the-clock exploration, their work expands from isolated predictions into long trajectories of reasoning, tool use, and feedback. Token efficiency therefore becomes important for scaling recursive self-improvement. We take an RSI-inspired approach at the harness layer, scaling auto-research loops across increasingly numerous and diverse environments for harness rollouts. At this scale, the process yields reusable improvements that transfer beyond their development setting, moving automated harness discovery toward production-level outcomes. Four mechanisms survive selection and form SoL-Pi, spanning action execution, context compaction, observation handling, and delegated reading. On the 51-task EdgeBench evaluation, SoL-Pi achieves performance comparable to Pi across GPT-5.6 Sol and Opus 5 while reducing recorded token traffic by 44.7-49.0% and API cost by about one third. In other words, estimated hourly savings are \$8.75-\$13.50 relative to native Codex and Claude Code harnesses, and \$4.36-\$5.71 relative to Pi.
Comments15 pages, 8 figures, 4 tables. Code: https://github.com/NVlabs/SoL-Pi . Project page: https://nvlabs.github.io/SoL-Pi/
