AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

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

Sicheng Zhang, Muhammad Muzammal Naseer, Binzhu Xie, Naufal Suryanto, Shi Qiu, Jamal Bentahar, NAVEED AKHTAR, Shah Mubarak

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Stage 1 reasoning-aware alignment continual pretraining on ReasonLite-42M (ablation)CLIP ViT-L/14 336 · COCO-5K I->TR@1: 58.0 -> 65.1Improved
add Stage 1 continual pretraining (ablation)CLIP ViT-L/14 336 · Urban-1K I->TR@1: 73.0 -> 83.0Improved
add Stage 2 explicit reasoning supervision after Stage 1 (CLIP-L) (ablation)ReasonCLIP L/14 224 · COCO-5K I->TR@1: 64.5 -> 59.9Hurt
add Stage 2 explicit reasoning supervision after Stage 1 (ablation)ReasonCLIP L/14 336 · RCLIP-V3-5K T->IR@1: 38.5 -> 42.8Improved
add Stage 2 for large SigLIP backbone (ablation)ReasonSigLIP So400m · RCLIP-V3-5K T->IR@1: 43.4 -> 49.5Improved
reasoning continual pretraining on compositional reasoning benchmarksCLIP B/32 · compositional reasoning avgaverage accuracy: 52.1 -> 57.7Improved
reasoning continual pretraining on SigLIP2 G/16SigLIP2 G/16 · SugarCrepeaccuracy: 83.0 -> 76.8Hurt
apply READ-CLIP fine-tuning on top of ReasonCLIP-B/32ReasonCLIP B/32 · compositional reasoning avgaverage accuracy: 64.1 -> 65.7Improved
replace CLIP with ReasonCLIP as LLaVA-NeXT visual towerLLaVA-NeXT (Qwen3-1.7B) · MMVPaccuracy: 58.7 -> 62.3Improved
replace CLIP with ReasonCLIP as visual towerLLaVA-NeXT (Qwen3-1.7B) · SciQAaccuracy: 73.1 -> 73.0No clear effect
stage 1 loss weighting: descriptive-heavy or reasoning-heavy instead of default schedule (ablation)ReasonCLIP Stage 1 · ablation on training configurationsoverallHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy