Pith. sign in

REVIEW 11 cited by

EVA-CLIP-18B: Scaling CLIP to 18 Billion Parameters

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2402.04252 v1 pith:WYGFUPKU submitted 2024-02-06 cs.CV

Quan Sun , Jinsheng Wang , Qiying Yu , Yufeng Cui , Fan Zhang , Xiaosong Zhang , Xinlong Wang This is my paper
classification cs.CV
keywords billionclipeva-clip-18bmodelmodelsscalingparametersavailable
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Scaling up contrastive language-image pretraining (CLIP) is critical for empowering both vision and multimodal models. We present EVA-CLIP-18B, the largest and most powerful open-source CLIP model to date, with 18-billion parameters. With only 6-billion training samples seen, EVA-CLIP-18B achieves an exceptional 80.7% zero-shot top-1 accuracy averaged across 27 widely recognized image classification benchmarks, outperforming its forerunner EVA-CLIP (5-billion parameters) and other open-source CLIP models by a large margin. Remarkably, we observe a consistent performance improvement with the model size scaling of EVA-CLIP, despite maintaining a constant training dataset of 2-billion image-text pairs from LAION-2B and COYO-700M. This dataset is openly available and much smaller than the in-house datasets (e.g., DFN-5B, WebLI-10B) employed in other state-of-the-art CLIP models. EVA-CLIP-18B demonstrates the potential of EVA-style weak-to-strong visual model scaling. With our model weights made publicly available, we hope to facilitate future research in vision and multimodal foundation models.

Discussion (0). Sign in to comment.

Forward citations

Cited by 11 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. UniHEAR: Unified Heterogeneous-Source Attentive Retrieval for Knowledge-Based Visual Question Answering

    cs.IR 2026-08 conditional novelty 6.0 of 10

    UniHEAR combines image-to-image and image-to-text candidate retrieval with source-aware attention reranking, improving Recall@1 over prior reranking methods on E-VQA and InfoSeek.

  2. MG$^2$-RAG: Multi-Granularity Graph for Multimodal Retrieval-Augmented Generation

    cs.IR 2026-04 unverdicted novelty 6.0 of 10

    MG²-RAG proposes a multi-granularity graph RAG framework that constructs hierarchical multimodal nodes via entity-driven visual grounding and performs structured retrieval, delivering SOTA results on four multimodal t...

  3. Explicit Logic Channel for Validation and Enhancement of MLLMs on Zero-Shot Tasks

    cs.AI 2026-03 unverdicted novelty 6.0 of 10

    Introduces Explicit Logic Channel (ELC) with LLM, VFM and probabilistic inference for validating, selecting and enhancing MLLMs on zero-shot tasks using Consistency Rate and cross-channel integration.

  4. Category-level Text-to-Image Retrieval Improved: Bridging the Domain Gap with Diffusion Models and Vision Encoders

    cs.CV 2025-08 conditional novelty 6.0 of 10

    Category-level retrieval improves when text queries are converted into multiple generated images, aggregated with a learned attention module, and fused with CLIP text similarity.

  5. ExpStar: Towards Automatic Commentary Generation for Multi-discipline Scientific Experiments

    cs.CV 2025-07 conditional novelty 6.0 of 10

    ExpStar, with a new 7,714-sample ExpInstruct dataset, generates step-level scientific experiment commentary including procedures, principles, and safety guidelines.

  6. Temporal Chain of Thought: Long-Video Understanding by Thinking in Frames

    cs.LG 2025-07 conditional novelty 6.0 of 10

    TCoT uses a single VLM to select question-relevant video frames from segments, then answers from that curated context, improving video QA accuracy across four benchmarks and three VLMs.

  7. mRAG: Elucidating the Design Space of Multi-modal Retrieval-Augmented Generation

    cs.AI 2025-05 conditional novelty 6.0 of 10

    A systematic empirical study finds that for multimodal RAG, EVA-CLIP retrieval, listwise LVLM reranking, and feeding only the top-ranked document works best, with a self-reflection agent adding further gains.

  8. Spa-VLM: Stealthy Poisoning Attacks on RAG-based VLM

    cs.CR 2025-05 conditional novelty 6.0 of 10

    Spa-VLM crafts paired adversarial images and misleading texts to poison RAG-based VLM knowledge bases, reaching attack success rates above 0.8 with just five injected entries.

  9. Magic-MM-Embedding: Towards Visual-Token-Efficient Universal Multimodal Embedding with MLLMs

    cs.CV 2026-02 conditional novelty 5.0 of 10

    Visual token compression (4x fewer tokens) plus a three-stage generative/contrastive/judge-curated training pipeline yields state-of-the-art MLLM-based retrieval accuracy at lower inference cost.

  10. Improve Multi-Modal Embedding Learning via Explicit Hard Negative Gradient Amplifying

    cs.CV 2025-05 conditional novelty 4.0 of 10

    A hard-negative gradient amplifier improves multimodal contrastive embedding training, achieving 72.5 average on MMEB, but it is a heuristic reweighting rather than a theoretical advance.

  11. Can Argus Judge Them All? Comparing VLMs Across Domains

    cs.IR 2025-06 reject novelty 2.0 of 10

    A VLM benchmark paper that proposes a cross-dataset consistency metric, but the abstract and body evaluate different model sets and the metric's bounds are incorrect.

Pith tools