Metraipolis › archive › dc_49be6224bcc2

SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness — arXiv:2609.20519

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.20519
License (as archived)
CC0-1.0
Archived
2026-10-10T03:22:09.439Z
SHA-256 of the archived text
1ad73f2954795f1c74b59b3513a7e1d080644b40707d1c85c4eda3e0fba3f995
Stable URL of this page
https://metraipolis.com/documento/dc_49be6224bcc2

How to cite

SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness — arXiv:2609.20519 — arXiv (en), https://arxiv.org/abs/2609.20519 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_49be6224bcc2 (Metraipolis archive dc_49be6224bcc2, sha256 1ad73f2954795f1c74b59b3513a7e1d080644b40707d1c85c4eda3e0fba3f995)
BibTeX
@misc{dc49be6224bcc2,
  title = {SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness — arXiv:2609.20519},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.20519},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_49be6224bcc2; sha256 1ad73f2954795f1c74b59b3513a7e1d080644b40707d1c85c4eda3e0fba3f995},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.20519. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

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.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.20519 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-10T03:04:13.747Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).