ChinaHeritaQA is a new bilingual VQA benchmark dataset with 2,279 images and 14,133 QA pairs for evaluating cultural reasoning abilities of VLMs on Chinese World Heritage sites across seven cognitive dimensions.
Title resolution pending
4 Pith papers cite this work, alongside 40 external citations. Polarity classification is still indexing.
4
Pith papers citing it
40
external citations · Crossref
years
2026 4representative citing papers
Cross-modal unlearning transfer in vision-language models is asymmetric, architecture-dependent, and shallow under typographic attacks; influence-guided block selection reduces the measured gap.
Text-conditioned sparse visual routing plus verifier-accepted-prefix rewards lengthens accepted draft prefixes and improves exact multimodal speculative decoding speedup over strong baselines.
Introduces self-captioning and a Multimodal Interaction Gate to amplify redundant multimodal interactions, reporting 38.3% reduction in visual-induced errors and 16.8% consistency improvement.
citing papers explorer
-
ChinaHeritaQA: A Culturally-Grounded Visual Question Answering Dataset for World Heritage Sites in China
ChinaHeritaQA is a new bilingual VQA benchmark dataset with 2,279 images and 14,133 QA pairs for evaluating cultural reasoning abilities of VLMs on Chinese World Heritage sites across seven cognitive dimensions.
-
One Modality to Forget Them All: Enhancing Cross-Modal Unlearning in Vision-Language Models
Cross-modal unlearning transfer in vision-language models is asymmetric, architecture-dependent, and shallow under typographic attacks; influence-guided block selection reduces the measured gap.
-
TIGER: Text-Conditioned Visual Gated Routing with Acceptance Alignment for Multimodal Speculative Decoding
Text-conditioned sparse visual routing plus verifier-accepted-prefix rewards lengthens accepted draft prefixes and improves exact multimodal speculative decoding speedup over strong baselines.
-
Self-Captioning Multimodal Interaction Tuning: Amplifying Exploitable Redundancies for Robust Vision Language Models
Introduces self-captioning and a Multimodal Interaction Gate to amplify redundant multimodal interactions, reporting 38.3% reduction in visual-induced errors and 16.8% consistency improvement.