← Search

Ethan J. Armand

2 accepted papers

2026

VideoNSA: Native Sparse Attention Scales Video Understanding

ICLR 2026poster

Video understanding in multimodal language models remains limited by context length: models often miss key transition frames and struggle to maintain coherence across long time scales. To address this, we adapt Native Sparse Attention (NSA) to video-language models. **Our method, VideoNSA, adapts Q…

Cited by 0SourcecodeScholar
2025

OverLayBench: A Benchmark for Layout-to-Image Generation with Dense Overlaps

NeurIPS 2025poster

Despite steady progress in layout-to-image generation, current methods still struggle with layouts containing significant overlap between bounding boxes. We identify two primary challenges: (1) large overlapping regions and (2) overlapping instances with minimal semantic distinction. Through both qu…

Cited by 0SourcecodeScholar