
提案された多言語VIFデータセットの構築方法とX-LLaVAモデルの効率的な多言語拡張フレームワークを示す。


coremsg

x-llava-optimizing-bilingual-large-vision-language-alignment-study


X-LLaVA: Optimizing Bilingual Large Vision-Language Alignment Study


title_rewrite


Proposing cost-effective methods for multilingual LMM training and dataset construction.


optimizing-bilingual-large-vision-language-alignment-study-at-seoul-national-university


Optimizing Bilingual Large Vision-Language Alignment Study at Seoul National University



Multimodal foundation models like CLIP show robustness under natural distribution shifts but fail to improve robustness under synthetic distribution shifts and adversarial attacks.


benchmarking-zero-shot-robustness-of-multimodal-foundation-models-a-comprehensive-evaluation


Benchmarking Zero-Shot Robustness of Multimodal Foundation Models: A Comprehensive Evaluation



GPT-4V shows promise as a generalist web agent with the SEEACT study, highlighting its potential for completing tasks on live websites.


exploring-gpt-4v-as-a-generalist-web-agent-seeact-study


Exploring GPT-4V as a Generalist Web Agent: SEEACT Study



Large Multimodal Models face distractibility from typographic attacks, but can be mitigated by providing more informative prompts.


Typographic-Attacks-in-Large-Multimodal-Models-Alleviating-with-Informative-Prompts


Typographic Attacks in Large Multimodal Models: Alleviating with Informative Prompts
