GAREN: A 890.9 FPS, 412 iter/s 3D Gaussian Splatting Rendering and Modeling Accelerator with Hybrid Sorting, Out of Range Filtering, and Gradient Skipping

Trident: A 28nm 1.83-to-9.74 pJ/bit Multi-Parameter FHE Processor with Unified Constant Geometry Datapath for Privacy Preserving Client-Server Computing

APT: Accelerating Diffusion Transformers via Attention Probability-Guided Pruning and Quantization

NetFED: Network-Aware Acceleration for Multi-Key Privacy-Preserving Federated Learning Aggregation

NOVA: Technology-Architecture Co-Design of Near-Memory Processing for Attention-SSM-MoE Hybrid LLM Inference

3DLS: A 3D Logic-Stacked Architecture for Disaggregated LLM Serving

Rethinking Token Reduction for Diffusion Models via Output-Similarity-Awareness

FastSTAR: Spatiotemporal Token Pruning for Efficient Autoregressive Video Synthesis

Adelia: A 4-nm LLM Processing Unit with Streamlined Dataflow and Dual-Mode Parallelism for Maximizing Hardware Efficiency

A 11.0 TOPS/W Diffusion Accelerator with Temporal Data Reuse for Real-Time Text-to-Motion Generation