A teacher-guided dense reward for GRPO distillation improves small-model tool-use accuracy and out-of-domain generalization on executable-tool benchmarks.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
MENTOR: Reinforcement Learning via Flexible Teacher-Optimized Rewards for Tool-Use Distillation
A teacher-guided dense reward for GRPO distillation improves small-model tool-use accuracy and out-of-domain generalization on executable-tool benchmarks.