InfoTree
Cross-source consensus on InfoTree from 1 sources and 5 claims.
1 sources · 5 claims
How it works
Benefits
Dosage & preparation
Interactions
Other
Highlighted claims
- InfoTree is presented as a training-time, budget-aware tree-search framework. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning
- InfoTree initializes from a root prompt, samples initial trajectories, computes entropy statistics, and expands frontier nodes using UUCB under a leaf budget. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning
- The main InfoTree configuration uses a 16-leaf training budget per prompt and a 32-leaf validation budget. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning
- InfoTree improved over flat GRPO across nine benchmarks by 2.5 to 11.2 points. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning
- InfoTree can be combined with DPS and prefix sharing for better results than InfoTree alone. — Maximizing Rollout Informativeness under a Fixed Budget: A Submodular View of Tree Search for Tool-Use Agentic Reinforcement Learning