Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

ReasonCLIP-58M: Visually Grounded Commonsense Reasoning Supervision for CLIP

About

CLIP and its variants are widely adopted visual backbones in multimodal systems, but their pretraining remains dominated by descriptive image-text alignment. As downstream applications increasingly demand visually grounded commonsense inference and compositional reasoning, it remains unclear whether CLIP-style encoders can support such reasoning without architectural changes. To address this, we present ReasonCLIP-58M, a continual pretraining framework that integrates large-scale reasoning supervision into CLIP-style models through our two-stage strategy, which progressively integrates reasoning signals while preserving descriptive alignment, followed by category-structured reasoning supervision. To support this framework, we construct two complementary datasets and a benchmark: ReasonLite-42M, with open-form, visually verifiable reasoning captions; ReasonPro-16M, with category-specific reasoning supervision; and RCLIP-Bench for diagnostic evaluation of visually grounded reasoning. We train a family of ReasonCLIP that improves visually grounded commonsense and compositional reasoning while also enhancing zero-shot retrieval performance. As a drop-in visual encoder for multimodal large language models such as LLaVA-NeXT, ReasonCLIP delivers consistent gains without additional inference cost, demonstrating that structured reasoning supervision enhances the expressive capacity of CLIP-style visual representations. All datasets, models, and training code are available at https://github.com/RISys-Lab/ReasonCLIP.

Sicheng Zhang, Muzammal Naseer, Binzhu Xie, Naufal Suryanto, Shi Qiu, Jamal Bentahar, Naveed Akhtar, Mubarak Shah• 2026

Related benchmarks

TaskDatasetResultRank
Text-to-Image RetrievalFlickr30k (test)
Recall@140.9
528
Compositional ReasoningSugarCrepe
Overall Accuracy88.5
95
Compositional ReasoningVALSE
Average Score79.2
65
Image-to-Text RetrievalCOCO 5K (test)
R@156.2
63
Commonsense ReasoningWinoGAViL
Accuracy (5-6)65.2
22
Compositional ReasoningWhatsUp
Accuracy51.3
21
Compositional ReasoningCREPE
Accuracy44.9
21
Image-to-Text Compositional ReasoningSugarCrepe++ ITT
Accuracy73.1
21
Text-to-Image Compositional ReasoningSugarCrepe++ TOT
Accuracy72.9
21
Image-to-Text RetrievalUrban1k (test)
R@178.6
9
Showing 10 of 14 rows

Other info

Follow for update