Blog - Adele Chinda

Multimodality, Diffusion

DraCo: Revolutionizing Text-to-Image Generation with Visual Chain-of-Thought

December 6, 2025

An in-depth analysis of how Draft-as-CoT achieves +8% improvement on GenEval through visual chain-of-thought planning for improved text-to-image generation.

Diffusion

NeuralRemaster: Phase-Preserving Diffusion for Structure-Aligned Generation

December 4, 2025

An in-depth exploration of phase-preserving diffusion mechanisms that maintain spatial structure while enabling flexible content generation for re-rendering and sim-to-real transfer.

Multimodality

ARM-Thinker: Agentic Multimodal Reward Models with Tool Use and Visual Reasoning

December 4, 2025

How agentic tool use and visual reasoning enable +16.2% improvement on reward modeling benchmarks for vision-language alignment.

Diffusion

Latent Diffusion for Natural Language Generation: Efficient Text in Learned Spaces

October 17, 2025

How latent diffusion models enable efficient and controllable text generation by operating diffusion in learned continuous latent spaces.

Diffusion

Encoder-Decoder Diffusion Language Models: Efficient Training and Inference

October 22, 2025

How encoder-decoder architectures optimize diffusion language models with 69% faster training and 345% faster inference while maintaining quality.