AI Paper Atlas

Research map / Video / Video generation

Neodragon: Mobile Video Generation Using Diffusion Transformer

Animesh Karnewar, Denis Korzhenkov, Ioannis Lelekas, Noor Fathima, Adil Karjauv, Mohsen Ghafoorian, Amir Habibian

ICLR 2026 · Video › Video generation · approach: Efficient video diffusion models

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace T5-XXL with distilled DT5 + ContextAdapter (Replace-Mode)Pyramidal-Flow · VBenchVBench total: 80.31 -> 79.64No clear effect
use LoRA adapters instead of MLP ContextAdapter (ablation)DT5 text encoder · VBenchVBench total: 79.64 -> 64.74Hurt
extend (cascade) CA with frozen CE instead of replacing CE (ablation)DT5 distillation · VBenchVBench total: 79.64 -> 79.16Hurt
add cosine distance loss to distillation (ablation)DT5 Replace-Mode · training stabilityconvergenceImproved
swap native decoder for distilled TinyAEHV decoderPyramidal-Flow · VBenchVBench total: 80.31 -> 80.25No clear effect
prune 6 of 24 MMDiT blocks, Stage-1 finetune onlyPyramidal-Flow MMDiT · VBenchVBench total: 80.31 -> 78.39Hurt
add Stage-2 full-teacher distillation after Stage-1 finetune (ablation)18-block MMDiT · VBenchVBench total: 78.39 -> 80.21Improved
prune 8 of 24 blocks with Stage-1 only (ablation)MMDiT · VBenchVBench total: 80.31 -> 74.59Hurt
prune blocks (24 -> 18) for latencyMMDiT on Hexagon NPU · latencyper-step latency: 1.15s -> 0.74sImproved
apply Stage-2 directly to pruned model without Stage-1 (ablation)pruned MMDiT · VBenchVBench totalHurt
use Next-Block mapping with data FM loss only plus feature loss instead of all losses/simple mapping (ablation)18-block MMDiT Stage-2 · VBenchVBench total: 80.21 -> 80.35Mixed
use Pyramidal DMD step distillation with 1-1-1 schedule vs native 1-1-1 (ablation)Pyramidal-Flow · VBenchVBench total: 59.62 -> 76.48Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy