aiqbench takeAI Products

NeoMME introduces a streamlined architecture for multimodal retrieval by eliminating redundant vision towers and causal decoders, directly addressing the parameter overhead inherent in current ColPali-style models. The approach achieves competitive nDCG@10 scores on ViDoRe v3 while significantly reducing compute requirements through a unified Transformer that processes raw image patches alongside text tokens. This architectural shift offers a more efficient path for deploying lightweight encoders in production environments where resource constraints are critical.

Source coveringMarkTechPost

What this isA model's reading of the article and related coverage, written for aiqbench. Not the publisher's words.

H Company Releases NeoMME: A Family of 260M and 800M Single-Tower Multimodal Encoders That Drop the Vision Tower and Causal Decoder

MarkTechPostSeptember 6, 20264 min read

We look at NeoMME, a family of 260M and 800M bidirectional encoders from H Company. Unlike ColPali-style retrievers, it processes multilingual text tokens and raw 32×32 image patches in a single Transformer, with no pretrained vision tower and no causal decoder. We cover the masked discrete-diffusion pretraining objective, the dual dense and late-interaction retrieval heads, and the ViDoRe v3 results where the 260M model reaches 0.523 nDCG@10. We also break down the 255× index compression, the 51.3 pages per second indexing throughput on one L40S, and the text-retrieval gaps the authors acknowledge. The post H Company Releases NeoMME: A Family of 260M and 800M Single-Tower Multimodal Encoders That Drop the Vision Tower and Causal Decoder appeared first on MarkTechPost.

Most visual document retrievers in production today are hand-me-downs. ColPali and the models that followed it take a generative vision-language model and repurpose it as an encoder. The result still carries a separately pretrained vision tower and a causal decoder that never generates a token. That is parameter and compute overhead for a task that only needs representations. H Company has released NeoMME, a family of 260M and 800M bidirectional encoders that drops…

Excerpt shown under the publisher's RSS terms · 73 words · full text and figures are on the publisher's site

Read at MarkTechPost ↗Ask about this article