Crimson AI NewsA CrimsonLingua Network service
EN ع
← Back to news
Research paper Hugging Face

DistilVDR: Compact 524M Visual Document Retriever Achieves Near-Teacher Accuracy via Dual-Student Distillation

AI By Crimson AI Hugging Face Papers 12 August 2026 · 00:00 7 views
Share: X Telegram

Hugging Face researchers introduce DistilVDR, a 524M-parameter visual document retriever distilled from an 8B teacher using cosine alignment without relevance labels, achieving 86.9% of teacher NDCG@5 with a 15.6x smaller index and 10x faster indexing.

DistilVDR: Compact 524M Visual Document Retriever Achieves Near-Teacher Accuracy via Dual-Student Distillation

Key points

Visual document retrieval (VDR) has been dominated by multi-billion-parameter models that are slow to index at full corpus scale and expensive to serve. Existing compression methods either train a smaller multi-vector encoder from scratch or distill only the query side, failing to produce a compact end-to-end single-vector retriever.

In a new paper, Hugging Face researchers present DistilVDR, a 524M-parameter end-to-end VDR system distilled bilaterally from a single 8B vision-language teacher using a pointwise cosine alignment loss. All supervision comes from the frozen teacher's embedding space, which was itself trained with relevance supervision, so the student objective requires no relevance labels, negative sampling, or contrastive terms.

The model mirrors VDR's text-query and image-document input asymmetry with an asymmetric encoder-only student that concentrates visual capacity on the document side (457M parameters) while keeping the query side at 70M parameters. Two variants share the same encoders and training but differ in the document encoder's visual-tile budget: DistilVDR-HiRes attains 61.74 average NDCG@5 on ViDoRe v1+v2+v3 (86.9% of the 8B teacher) and leads every reproduced sub-1B baseline on the high-resolution-sensitive v3 benchmark, while DistilVDR-Fast attains 59.98 at a 3x smaller visual-token budget.

Both variants store one million documents in a 15.6x smaller index than the strongest sub-1B multi-vector baseline and index the corpus an order of magnitude faster. The code is available on GitHub, and training details and models are released on the NanoVDR Hugging Face space.

ModelParametersAvg NDCG@5 (ViDoRe v1+v2+v3)% of TeacherIndex Size per 1M docsIndexing Speed
DistilVDR-HiRes524M (457M doc + 70M query)61.7486.9%16.4 GB37 pages/s
DistilVDR-Fast524M (457M doc + 70M query)59.9884.4%16.4 GB99 pages/s
8B Teacher8B~71.0 (estimated)100%256 GB (multi-vector)5.4 pages/s
Source
Hugging Face · Hugging Face Papers
Related news
Research paper
Hugging Face 31 Aug 2026

Hugging Face Unveils StepGuard: Step-Level Guardrails for Safer AI Agents

StepGuard, a new step-level guard model from Hugging Face, audits agent actions before execution, reducing attack success rates by...

0
Research paper
Hugging Face 31 Aug 2026

Hugging Face Researchers Unveil ABot-Recon for Stable Long-Horizon 3D Reconstruction

ABot-Recon, a new streaming 3D reconstruction model from Hugging Face, achieves stable long-horizon performance using only local t...

0
Research paper
Hugging Face 31 Aug 2026

ContextPilot: Teaching Agents Proactive Context Management via Fine-Grained RL

Hugging Face researchers introduce ContextPilot, a framework that enhances long-horizon agent reasoning by expanding context-editi...

1