[KongchangAI]
Modelpplx-embed-v2-late-0.6b

pplx-embed-v2-late-0.6b

A lightweight 0.6B-parameter version of the pplx-embed-v2-late series, aimed at low-latency query scenarios. Its performance on visual document retrieval tasks is comparable to that of models with several times as many parameters.

Timeline (last 90 days)

Oct 22

Because the two models share the same embedding space, they support an asymmetric deployment scheme in which the 9B model builds the document index and the 0.6B model handles online queries.

Unverified50%
Oct 22

The 0.6B version can compete with models several times its parameter count on visual document retrieval tasks.

Unverified50%

All Facts (2)

Source Articles