Factored Inference Hits 85.5%, Pruning Takes 2.5ms

Also Worth Noting

• Breaking the Synthetic-Real Domain Shortcut for Training-Free Generative Replay-based Class Incremental Learning score 4
  关键词(1): text-to-image; 顶会接收: ICML
• Structured Redundancy Modeling for Efficient Visual Token Pruning in High-Resolution MLLMs score 4
  关键词(3): compression, pruning, latency; 顶会接收: ECCV
• Similarity Is Not Logic: Factored Inference for Dual-Encoder Vision-Language Models score 4
  关键词(2): fine-tuning, vision-language; 顶会接收: ICML
• Stabilizing Deep Reconstruction Operators with Contractive Anchoring score 4
  关键词(1): lightweight; 顶会接收: ECCV
• UltraViT: Latency-Optimized On-device Vision Encoder for Large Vision-Language Models score 4
  关键词(6): distillation, deployment, latency, edge, pre-training; 顶会接收: ECCV
• Trainable Nonexpansive Denoisers for Contractive Image Reconstruction score 3
  顶会接收: ICML