InterleaveThinker: Reinforcing Agentic Interleaved Generation

View on Reddit

Paper proposes InterleaveThinker, a method that uses reinforcement learning to improve agentic interleaved generation in image models, enhancing photorealism and instruction following. Code and paper are open source.