6 papers
Periodic RoPE for Infinite Context LLMs
Simin Huo
The ability to process ultra-long contexts is crucial for large language models (LLMs) to perform long-horizon tasks. While recent efforts have extended context windows to 1M and b…
TTF: Temporal Token Fusion for Efficient Video-Language Model
Simin Huo, Ning LI
Video-language models (VLMs) face rapid inference costs as visual token counts scale with video length. For example, 32 frames at resolution already yield >8,000 v…
MaMe & MaRe: Matrix-Based Token Merging and Restoration for Efficient Visual Perception and Synthesis
Simin Huo, Ning Li
Token compression is crucial for mitigating the quadratic complexity of self-attention mechanisms in Vision Transformers (ViTs), which often involve numerous input tokens. Existing…
Iwin Transformer: Hierarchical Vision Transformer using Interleaved Windows
Simin Huo, Ning Li
Vision Transformers (ViTs) face two limitations: the rigid resolution dependency of positional embeddings, which complicates cross-resolution fine-tuning, and the quadratic complex…
ULU: A Unified Activation Function
Simin Huo
We propose \textbf{ULU}, a novel non-monotonic, piecewise activation function defined as , where . ULU treats po…
Single Domain Generalization for Few-Shot Counting via Universal Representation Matching
Xianing Chen, Si Huo, Borui Jiang +2
Few-shot counting estimates the number of target objects in an image using only a few annotated exemplars. However, domain shift severely hinders existing methods to generalize to…