tencent/EVIE-8B and EVIE-4.5B (High-Capacity Visual Document Retrieval)
腾讯推出了EVIE-8B和EVIE-4.5B模型，专为高容量视觉文档检索设计。这些模型已通过138项任务的广泛验证，包括ViDoRe V1、V2、V3和JinaVDR，并采用了nDCG、Recall、MAP和MRR u/1这四种标准度量系列。其核心特点是EVIE-ARD蒸馏配方，这是一种保留锚点、感知容量的关系蒸馏方法，能够从8B教师模型中重现完整的学生训练过程。
- 发布
- 2026年9月7日 09:47
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 正常
时间以 UTC 显示
更多信息
https://preview.redd.it/3p6234jzk2oh1.png?width=900&format=png&auto=webp&s=b9eab3351d6c6dbd4d5b3fd677a6c40b57f18167
- SOTA Retrieval Performance: 66.75 nDCG@10 on ViDoRe V3, delivering industry-leading visual document retrieval accuracy.
- High-Capacity 4096D Representations: Full per-token multi-vector embeddings preserving fine-grained layout, typography, charts, and table structures.
- Teacher Foundation: Provides capacity-aware relation and margin distillation targets for the lightweight EVIE-4.5B Prefix-MRL model.
- Multi-Benchmark 138-Task Coverage: Thoroughly validated across 138 tasks (ViDoRe V1, V2, V3, and JinaVDR) across 4 standard metric families (nDCG, Recall, MAP, MRR u/1 ).
- Top-Tier Benchmark Performance: 66.75 on ViDoRe V3 for EVIE-8B and 66.02 for EVIE-4.5B with single-projection Prefix-MRL.
- ⚡ Prefix-MRL Elasticity: Single 2048D linear projection. Freely truncate at runtime into ${64, 128, 256, 512, 1024, 2048}$ dimensions without separate models.
- 📦 Ultra-Compact Index (HAC): Training-free Hierarchical Agglomerative Clustering compresses token counts from ~750 down to 32 vectors/page, slashing index storage to 3.81 GiB per million pages.
- 🌐 138 Multilingual Tasks Evaluated: Thoroughly evaluated across ViDoRe V1, V2, V3, and JinaVDR across 4 metric families (nDCG, Recall, MAP, MRR u/1 ).
- 🔬 EVIE-ARD Distillation Recipe: Anchor-preserving, capacity-aware relation distillation reproducing full student training from the 8B teacher.