SharpMoE is a plug-and-play post-training method that uses clean latent features and a trajectory routing loss to enable accurate saliency-based routing in diffusion MoE models for improved visual generation.
hub
Scaling Diffusion Transformers to 16 Billion Parameters
15 Pith papers cite this work. Polarity classification is still indexing.
15
Pith papers citing it
hub tools
citation-role summary
background 2
method 1
citation-polarity summary
verdicts
UNVERDICTED 15representative citing papers
DiT-Reward converts pretrained DiT models into reward predictors that outperform HPSv3 on four benchmarks while providing 1.65x inference speedup.
A kinematic-to-visual lifting paradigm combined with hierarchically routed control generates action-conditioned surgical videos with better faithfulness, fidelity, and efficiency.
Large loss spikes in SGD are polynomially likely and serve as the dominant mechanism for escaping sharp minima toward flatter solutions in the NTK regime.
InstructMoLE replaces per-token routing with instruction-guided global routing for mixture-of-low-rank-experts in diffusion transformers and adds an output-space orthogonality loss to improve multi-conditional image generation.
MADreMIA amplifies membership inference signals by showing that memorized samples maintain higher coherence and slower degradation in chained regeneration trajectories than non-members.
WiSP achieves up to 1.95x decode throughput on low-resource MoE serving by dynamically paging reused experts and using MV-WSA to allocate VRAM between experts and KV cache, with the offline policy performing well on both prefill and decode.
MoECodec replaces FFN layers with token-wise MoE plus stable routing and GShMLP experts to support multiple downstream tasks in a single image compression model.
CoInteract adds a human-aware mixture-of-experts and spatially-structured co-generation to a diffusion transformer to synthesize videos with stable structures and physically plausible human-object contacts.
VersaVogue unifies garment generation and virtual dressing via trait-routing attention with mixture-of-experts and an automated multi-perspective preference optimization pipeline that uses DPO without human labels.
TripoSG generates high-fidelity 3D meshes from input images via a large-scale rectified flow transformer and hybrid-trained 3D VAE on a custom 2-million-sample dataset, claiming state-of-the-art fidelity and generalization.
GenEraser proposes MC-MoE with bipartite text guidance, LD-CFG fusion, and a decoupled locator-preserver architecture for generalizable video object and effect removal, claiming 2.16 dB and 1.44 dB gains on ROSE and VOR-Eval benchmarks.
Diagnoses five failure modes in Token-Choice MoE routing for visual diffusion transformers and proposes the Functional Redundancy Hypothesis to explain selective deadlock.
FocusDiT masks non-critical query tokens before they enter the FFN in DiT models, directing capacity toward complex visual details and reporting improved text-to-image results.
Mamoda2.5 is a 25B-parameter DiT-MoE unified AR-Diffusion model that reaches top video generation and editing benchmarks with 4-step inference up to 95.9x faster than baselines.
citing papers explorer
-
MoECodec: Image Compression for joint human and machine perception via Mixture-of-Experts
MoECodec replaces FFN layers with token-wise MoE plus stable routing and GShMLP experts to support multiple downstream tasks in a single image compression model.