SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness
SoL-Pi scales auto-research loops to produce an agent harness matching Pi while cutting token traffic 44.7-49.0% and API cost by a third.
The paper takes an RSI-inspired approach at the harness layer, scaling auto-research loops across increasingly numerous and diverse environments to yield reusable improvements that transfer beyond their development setting. Four mechanisms survived selection to 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. Estimated hourly savings are $8.75-13.50 relative to native Codex and Claude Code harnesses.
- Auto-research loops yield reusable harness improvements that transfer across settings
- Four mechanisms span action execution, context compaction, observation handling, delegated reading
- Matches Pi on 51-task EdgeBench with 44.7-49.0% less token traffic
- Cuts API cost by about one third versus existing coding harnesses
- Products
- SoL-PiPiCodexClaude Code
- AI models
- GPT-5.6 SolOpus 5
From the original article
1 min read at huggingface.co
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.
By Haozhe Liu, Tian Ye, Sensen Gao · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.