AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

SOCO: Benchmarking Semantic Object Correspondence in Vision Foundation Models

Olaf Dünkel, Basavaraj Sunagad, Haoran Wang, David Hoffmann, Christian Theobalt, Adam Kortylewski

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Instruction-tuned multimodal assistants · 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
require object-relative identity (SOC) instead of concept matching only (CC)vision foundation models (all evaluated) · SOCOPCK@0.1: DINOv2 78.9 -> 60.4Hurt
move from within-category SOC to cross-category SOCDINOv2 · SOCOPCK@0.1: 60.4 -> 54.9Hurt
dense self-supervised objective (DINOv2) versus global alignment (CLIP)vision encoders · SOCO SOCPCK@0.1: 16.2 -> 60.4Improved
scale DINOv2 to DINOv3DINO · SOCO SOCPCK@0.1: SOC 60.4 -> 55.5; SOC-geo 71.9 -> 76.0Mixed
use reconstruction objective (MAE) instead of dense SSLvision encoders · SOCO SOCPCK@0.1: DINOv2 60.4 -> MAE 9.4Hurt
supply keypoint text description only instead of visual markerQwen3-VL-8B · SOCO LVLMCircularEval accuracy: 34.2 -> 54.0Improved
add description to visual queryQwen3-VL-8B · SOCO LVLMCircularEval accuracy: 34.2 -> 30.8Hurt
add description to visual queryGPT4o · SOCO LVLMCircularEval accuracy: 30.2 -> 30.9No clear effect
use DINOv2 instead of best LVLM in 4-choice protocolDINOv2 · SOCO LVLMCircularEval accuracy: 54.0 (best LVLM) -> 81.0Improved
scale Qwen3-VL from 4B to 8B and compare with Qwen2.5-VL-7BQwen · SOCO LVLM (Vis.)CircularEval accuracy: Qwen2.5-VL-7B 19.4 -> Qwen3-VL-8B 34.2Improved
train a linear probe on features instead of nearest-neighbor matching (ablation)vision encoders · SOCO SOCPCK@0.1: DINOv2 60.4 -> 69.8Improved
use window soft-argmax evaluation (ablation)vision encoders · SOCO SOCPCK@0.1: DINOv2 60.4 -> 62.7Improved

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