Thoughts on generative models, video quality, and the intersection of perception and computation.
Full PhD dissertation defense slides covering HDR-Q (MLLM for HDR VQA), LumaFlux (inverse tone mapping with diffusion transformers), and Rectified-CFG++ (geometry-aware guidance for flow models).
View slides →Presentation slides for HDR-Q (CVPR 2026) — the first multimodal LLM for HDR video quality assessment, featuring HAPO (HDR-Aware Policy Optimization) with contrastive KL, dual-entropy regularization, and SigLIP-2 HDR-aware encoding.
View slides →Presentation slides for LumaFlux — SDR-to-HDR inverse tone mapping using Flux 12B with Physically-Guided Adaptation (PGA), Perceptual Cross-Modulation (PCM), and Rational Quadratic Spline (RQS) decoder with only ~17M trainable parameters.
View slides →Presentation slides for Rectified-CFG++ (NeurIPS 2025) — a predictor-corrector guidance method that fixes CFG artifacts on flow models like Flux, SD3, and Lumina-Next with theoretical guarantees and zero extra training cost.
View slides →A long revision notebook for Research Scientist / GenAI loops focused on image generation, perceptual quality, and video processing: diffusion and flow models, transformer internals (attention variants, RoPE, KV cache), the text-to-image design space, evaluation metrics, color and HDR, classical CV, RL alignment, and the coding tier — all in one place.
Read more →An intuition-first deep dive into Rectified Flow with the core equations, geometric interpretation of trajectory straightness, and minimal PyTorch-style code snippets for training and sampling.
Read more →Why measuring quality in high dynamic range videos is fundamentally different from SDR, and what makes it one of the most challenging problems in perceptual quality research.
Read more →Why classifier-free guidance breaks on flow models, and how a geometry-aware predictor-corrector fixes it — with a deep dive into text rendering quality and the adaptive schedule that makes it work.
Read more →