Adaptive Budget Allocator
Cross-source consensus on Adaptive Budget Allocator from 1 sources and 5 claims.
1 sources · 5 claims
Uses
How it works
Benefits
Risks & contraindications
Highlighted claims
- ABA triggers only for uniform-outcome trees when predicted rescue probability exceeds 0.5. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning
- ABA handles uniform-outcome trees after the initial tree construction phase. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning
- ABA uses prompt and tree summaries to predict whether a high-temperature extension will create another outcome category. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning
- ABA increased mixed-outcome ratio and reduced uniform-fail ratio, with modest increases in leaves and wall-clock time. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning
- ABA performance degraded under distribution shift, but retraining every 100 steps restored ROC-AUC and lift. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning