CVPR 2026 accepted papers on Reinforcement learning
57 of the 4,042 papers accepted at CVPR 2026 (1.4%) are on Reinforcement learning, compared with 0.3% at CVPR 2025. They fall into 3 topics.
See CVPR 2026 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Visual and representation RL · 34 papers
- CUBic: Coordinated Unified Bimanual Perception and Control Framework
- Dual-Agent Reinforcement Learning for Adaptive and Cost-Aware Visual-Inertial Odometry
- Efficient Hybrid SE(3)-Equivariant Visuomotor Flow Policy via Spherical Harmonics for Robot Manipulation
- Explore with Long-term Memory: A Benchmark and Multimodal LLM-based Reinforcement Learning Framework for Embodied Exploration
- Goal-Driven Reward by Video Diffusion Models for Reinforcement Learning
- InternData-A1: Pioneering High-Fidelity Synthetic Data for Pre-training Generalist Policy
- InterPrior: Scaling Generative Control for Physics-Based Human-Object Interactions
- Learning to Assist: Physics-Grounded Human-Human Control via Multi-Agent Reinforcement Learning
- Local Motion Matters: A Deconstruct-Recompose Paradigm for Reinforcement Learning Pre-training from Videos
- MaskAdapt: Learning Flexible Motion Adaptation via Mask-Invariant Prior for Physics-Based Characters
- Opening the Sim-to-Real Door for Humanoid Pixel-to-Action Policy Transfer
- Planning in 8 Tokens: A Compact Discrete Tokenizer for Latent World Model
and 22 more in the interactive map.
RL from feedback · 22 papers
- Do Less, Achieve More: Do We Need Every-Step Optimization for RL Fine-tuning of Diffusion Models?
- Dynamic Important Example Mining for Reinforcement Finetuning
- Expand and Prune: Maximizing Trajectory Diversity for Effective GRPO in Generative Models
- Fine-Grained GRPO for Precise Preference Alignment in Flow Models
- From Exploration to Exploitation: A Two-Stage Entropy RLVR Approach for Noise-Tolerant MLLM Training
- GDRO: Group-level Reward Post-training Suitable for Diffusion Models
- GTR-Turbo: Merged Checkpoint is Secretly a Free Teacher for Agentic VLM Training
- MaskFocus: Focusing Policy Optimization on Critical Steps for Masked Image Generation
- MedLoc-R1: Performance-Aware Curriculum Reward Scheduling for GRPO-Based Medical Visual Grounding
- MM-ReCoder: Advancing Chart-to-Code Generation with Reinforcement Learning and Self-Correction
- Neighbor GRPO: Contrastive ODE Policy Optimization Aligns Flow Models
- OraPO: Oracle-educated Reinforcement Learning for Data-efficient and Factual Radiology Report Generation
and 10 more in the interactive map.