of wall-clock spent on rollout generation in agentic/reasoning RL post-training
~80%observed
| Value kind | observed — Reported measurements, counts, and specifications keep the precision and scope stated by their source; an exact specification is not treated as a range. |
|---|---|
| As of | 2026 |
| Source | 2025–2026 RL-systems papers (ROLL Flash, ROLLART) · Appendix A.7.4, page 36: rollout generation consumes 70–80% of total training time for the cited long-context agentic-RL configuration |
| Review | checking…review by 2026-09-09 · standard cadence |
| Recorded changes | last 2026-07-10 · 3 revisions tracked |
| Claim id | of-wall-clock-spent-on-rollout-generation-in |
Where the guide uses it
← Full numbers register — every date-stamped figure in the guide, with revision history.