NeoMME: single-tower multimodal encoder leads sub-800M models on ViDoRe v3 visual retrieval
_reachsumit · x · 2026-09-03
NeoMME is a family of 260M/800M bidirectional encoders processing multilingual text and raw image patches in one Transformer, pretrained from scratch with a masked discrete-diffusion objective and 16K-token context. Fine-tuned retrievers hit 0.523 (260M) and 0.556 (800M) nDCG@10 on ViDoRe v3, beating all evaluated models under 800M parameters, with 2x encoding throughput of ColModernVBERT on an L40S. Weights are on Hugging Face.
More from Multimodal
- JSON prompting with MOSS Image 5B: bbox coordinates for exact multi-room layouts — cocktailpeanut · 2026-10-21
- MOSS Image 5B lands in Maestro with raw JSON prompting for precise layout control — cocktailpeanut · 2026-10-21
- FastVideo open-sources MiniMax FastH3 Omni with training code, 12x faster on Blackwell — multimodalart · 2026-10-21
- Megalith 008: stunning AI-generated colossal architecture series — azed_ai · 2026-10-21
- Higgsfield offers $25K prizes for open-source ComfyUI rebuilds of its apps — MuziqueComfyUI · 2026-10-21
- Stunning AI-generated elderly portrait blurs the line with real photography — LudovicCreator · 2026-10-21