SOLE-R1: Video-Language Reasoning as the Sole Reward for On-Robot Reinforcement Learning
This addresses the challenge of enabling robots to learn manipulation tasks without ground-truth rewards or demonstrations, though it is incremental as it builds on existing vision-language model capabilities.
The paper tackled the problem of using vision-language models as reward signals in reinforcement learning for robots, which often fail due to partial observability and distribution shift, by introducing SOLE-R1, a video-language reasoning model that serves as the sole reward for online RL, enabling zero-shot learning on 24 unseen tasks and outperforming strong models like GPT-5 and Gemini-3-Pro.
Vision-language models (VLMs) have shown impressive capabilities across diverse tasks, motivating efforts to leverage these models to supervise robot learning. However, when used as evaluators in reinforcement learning (RL), today's strongest models often fail under partial observability and distribution shift, enabling policies to exploit perceptual errors rather than solve the task. To address this limitation, we introduce SOLE-R1 (Self-Observing LEarner), a video-language reasoning model explicitly designed to serve as the sole reward signal for online RL. Given only raw video observations and a natural-language goal, SOLE-R1 performs per-timestep spatiotemporal chain-of-thought (CoT) reasoning and produces dense estimates of task progress that can be used directly as rewards. To train SOLE-R1, we develop a large-scale video trajectory and reasoning synthesis pipeline that generates temporally grounded CoT traces aligned with continuous progress supervision. This data is combined with foundational spatial and multi-frame temporal reasoning, and used to train the model with a hybrid framework that couples supervised fine-tuning with RL from verifiable rewards. Across four different simulation environments and a real-robot setting, SOLE-R1 enables zero-shot online RL from random initialization: robots learn previously unseen manipulation tasks without ground-truth rewards, success indicators, demonstrations, or task-specific tuning. SOLE-R1 succeeds on 24 unseen tasks and substantially outperforms strong vision-language rewarders, including GPT-5 and Gemini-3-Pro, while exhibiting markedly greater robustness to reward hacking.