demystifying-rl-tool-agents

Comprehensive recipe for RL-training tool-using agents spanning reward design, data synthesis, model scaling, and algorithm selection. Seven ranked findings: scale-dependent rewards (curriculum for 1.5B–3B; dense for 7B), semi-sparse 'Macro' rewards balance specialization/transfer, 1K-sample sweet spot with 4:3:3 difficulty mix. Achieves SOTA on TravelPlann…

Independent GitHub snapshot · Automated classification · Original documentation governs