AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

ECC: Encoder-Centric Corruption for Fine-Grained Vision in VLMs

Hyesong Choi, Daeun Kim, Sungmin Cha, Kwang Moo Yi, Dongbo Min

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
use Conjunctive instead of Substitutive Corruption (ablation)noise-based C2R pretraining · FGVC and recognition benchmarkstransfer accuracyImproved
move corruption and reconstruction into the encoder (Encoder-style) vs Decoder-style (ablation)naive noise-based C2R · recognition and FGVCtransfer accuracyImproved
inject noise at feature level (blocks 2,4,6) instead of pixel level (block 0) (ablation)ECC encoder · recognition taskstransfer accuracyImproved
add disruption loss to separate mask reconstruction and denoising (ablation)ECC · fine-grained and standard recognitionaccuracyImproved
add noise-based denoising to MIM (DiffMAE, MaskDiT vs MAE)MAE · FGVC (CUB)top-1 accuracy: 79.14 -> 78.50Hurt
use ECC instead of MAEMAE · Aircrafttop-1 accuracy: 81.56 -> 84.42Improved
use ECC instead of MAEMAE · COCO (mask)AP: 41.74 -> 47.90Improved
apply ECC to CLIP vision encoderCLIP ViT-B/16 · fine-tuningaccuracy: 83.35 -> 84.83Improved
apply ECC to CLIP vision encoderCLIP · visual groundingattention entropy (bits): 4.2 -> 3.5Improved
apply ECC to LLaVA-1.5 vision encoderLLaVA-1.5 · object hallucinationCHAIRs: 48.4 -> 46.5Improved
apply ECC to LLaVA-1.5 vision encoderLLaVA-1.5 · Flickr30KR@1: 69.1 -> 70.2Improved

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