AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Corvid: Improving Multimodal Large Language Models Towards Chain-of-Thought Reasoning

Jingjing Jiang, Chao Ma, Xurui Song, Hanwang Zhang, Jun Luo

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 10 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 gate attention in GateMixer (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 54.1 -> 55.6Improved
add L_CReg contrastive alignment loss in pretraining (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 55.0 -> 55.6Improved
add learnable prefix token embeddings (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 54.8 -> 55.6Improved
use GateMixer instead of MLP connector (FC_GELU_FC, 2xFC_GELU_FC) (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 54.0 -> 55.6; 53.7 -> 55.6Improved
train on high-quality CoT data instead of direct-answer data (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 45.7 -> 55.6Improved
use refined CoTs instead of raw rationales (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 51.2 -> 55.6Improved
use Corvid-1M instead of LLaVA-665K, Cambrian-1M or Ovis-1M (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 44.1, 51.9, 53.8 -> 55.6Improved
increase SFT data diversity (CoT-only to Corvid-1M mix with direct and OCR data) (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 48.0 -> 55.6Improved
add inference-time self-verification over always-CoT inference (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 49.7 -> 55.6Improved
add inference-time self-verification over always-CoT inference (ablation)Corvid-o1 · 9 reasoning benchmarksaverage accuracy: 56.2 -> 59.9Improved
use hybrid SigLIP+ConvNeXt vision encoder instead of one encoder (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 54.7 -> 55.6; 52.3 -> 55.6Improved
raise self-verification weight alpha from 0.0 to 0.7 (ablation)Corvid-base · 9 reasoning benchmarksaverage accuracy: 48.6 -> 55.6Improved

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