CoCo: Code as CoT for Text-to-Image Preview and Rare Concept Generation

Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• CoCo (Code-as-CoT) represents the reasoning process as executable code for explicit and verifiable intermediate planning in image generation.method 95%
• CoCo generates executable code to specify the structural layout of a scene, which is then executed to render a deterministic draft image.method 95%
• CoCo refines draft images through fine-grained image editing to produce the final high-fidelity result.method 95%
• CoCo-10K is a curated dataset containing structured draft-final image pairs designed to train structured draft construction and corrective visual refinement.method 95%
• CoCo achieves improvements of +68.83% on StructT2IBench compared to direct generation.result 95%
• CoCo achieves improvements of +54.8% on OneIG-Bench compared to direct generation.result 95%
• CoCo achieves improvements of +41.23% on LongText-Bench compared to direct generation.result 95%
• Executable code is an effective and reliable reasoning paradigm for precise, controllable, and structured text-to-image generation.technical 95%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card