Meituan LongCat's FVQ makes discrete visual tokens match continuous representations
meituan-longcat · hf · 2026-10-28
A paper from Meituan LongCat with HUST, SJTU, PKU and Tsinghua revisits the assumption that quantization error inherently causes semantic loss in discrete visual tokens. The authors find quantization error unreliable—discrete representations with larger errors can achieve better understanding—and identify effective bit budget as the key determinant. Existing quantizers saturate as codebooks grow, so they propose Factorized Vector Quantization (FVQ), which quantizes complementary fixed-dimensional views of visual features with symmetric parallel codebooks, continuing to improve at larger bit budgets and ultimately matching continuous visual representations in multimodal understanding.
More from Research
- UltraVBERT: a 247M-parameter vision-language encoder for visual document retrieval — kalyan_kpl · 2026-10-28
- Anders Sandberg on a wild 48 hours: group classification formalized, Kakeya, RCE, AI incident — anderssandberg · 2026-10-28
- EvidenceConnect boosts multimodal perception without zooming, sweeping 8 benchmarks — inclusionAI · 2026-10-28
- What 200 years of philology reveals about LLM confabulation's structural root — vasilisvj · 2026-10-28
- Map-based localization hits 1.9cm RMSE across EuRoC flights in visloc-rs — rsasaki0109 · 2026-10-28
- BeeNara v0.4: 332MB local doc sorter hits 91.5% on unseen folders — razer_psycho · 2026-10-28