Adobe and UCLA Release Tev-Gym: 1,170 Long-Horizon Environments for Agentic RL

Adobe Research and UCLA introduced Tev-Gym, a collection of 1,170 verifiable long-horizon RL environments with dense progress rewards, addressing data scarcity and sparse rewards for open-source models. A 27B agent trained on it achieves performance comparable to frontier models.

2026-10-14 ~ 2026-10-15 · 2 related posts

• Tev-Gym: 1,170 verified long-horizon agent environments with dense progress rewards — Wenbo Hu · 2026-10-14
• Tev-Gym: 1,170 dense-reward RL environments push long-horizon agentic training — RulinShao · 2026-10-15