trends.barnas.net

method · first seen 15 Oct 2026

rollout truncation

Stopping generation early during rollouts to save computation in reinforcement or distillation training.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

on-policy distillationinverse propensity weighting