method · first seen 14 Oct 2026
decomposable verifiable rewards
Reinforcement learning rewards split into verifiable intermediate checkpoints that credit partial task completion.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Explorer-Zero: Improving Self-Play for Long-Horizon Tasks with Decomposable Verifiable RewardsPapers · 14 Oct 2026 · Tianyu Liu, Davide Belli, Bence Major, Arash Behboodi, Tribhuvanesh Orekondy
Latest sources
- Explorer-Zero: Improving Self-Play for Long-Horizon Tasks with Decomposable Verifiable RewardsPapers · 14 Oct 2026 · Tianyu Liu, Davide Belli, Bence Major, Arash Behboodi, Tribhuvanesh Orekondy