David-GRPO improves low-budget RL training for multi-hop QA agents by bootstrapping expert trajectories and converting on-policy partial successes into evidence-coverage signals that increase retrieval depth.
Now, I need to check if Laleli is part of the same neighborhood
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Can David Beat Goliath? On Multi-Hop Reasoning with Resource-Constrained Agents
David-GRPO improves low-budget RL training for multi-hop QA agents by bootstrapping expert trajectories and converting on-policy partial successes into evidence-coverage signals that increase retrieval depth.