AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Light-VQA+: A Video Quality Assessment Model for Exposure Correction with Vision-Language Guidance

Xunchu Zhou, Xiaohong Liu, Yunlong Dong, Tengchuan Kou, Yixuan Gao, Zicheng Zhang, Chunyi Li, Haoning Wu, Guangtao Zhai

IJCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis · 12 citations (OpenAlex)

Read the paper

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 handcrafted brightness and noise features (ablation)Light-VQA+ variant (MLP, no HVS weights) · VEC-QASRCC: 0.8614 -> 0.8712Improved
replace handcrafted features with CLIP local brightness/noise and Lv4 consistency features (ablation)Light-VQA+ variant (MLP, no HVS weights) · VEC-QASRCC: 0.8712 -> 0.9010Improved
add CLIP global features instead of local (ablation)Light-VQA+ variant (MLP, no HVS weights) · VEC-QASRCC: 0.9010 -> 0.9021No clear effect
add cross-attention fusion instead of MLP (ablation)Light-VQA+ · VEC-QASRCC: 0.9010 -> 0.9177Improved
add HVS-aligned clip weights instead of simple average (ablation)Light-VQA+ · VEC-QASRCC: 0.9056 -> 0.9177Improved
add HVS weights (ablation)Light-VQA+ · OEVR-QASRCC: 0.7215 -> 0.7462Improved
add cross-attention fusion (ablation)Light-VQA+ · LLVE-QASRCC: 0.9317 -> 0.9428Improved
use Light-VQA+ instead of Light-VQAVQA model · OEVR-QASRCC: 0.5991 -> 0.7407Improved
use Light-VQA+ instead of Light-VQAVQA model · VEC-QASRCC: 0.8712 -> 0.9121Improved
use Light-VQA+ instead of Light-VQA on unseen datasetsVQA model · Mixed-B cross-datasetSRCC: 0.6358 -> 0.7328Improved
fine-tune FEC-Net with Light-VQA+ in the lossFEC-Net · over-exposed videosLight-VQA+ score: 69.6 -> 74.2Improved
fine-tune FEC-Net with Light-VQA+ in the lossFEC-Net · low-light videosLight-VQA+ score: 59.8 -> 64.0Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 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