Odysseus: Scaling VLMs to 100+ Turn Decision-Making in Games via Reinforcement Learning

Benchmark Model Rank Results
multimodal-reasoning-on-mathvisionOdysseus (RL on SFT)–Accuracy: 53.52
multimodal-reasoning-on-mmmuOdysseus (RL on SFT)–Accuracy: 70.77
visual-question-answering-vqa-on-realworldqaOdysseus (RL on SFT)–Accuracy: 71.11