正经科研讯息 [长期更新]

[yaoyz96]

创新更多发生在科研网络的边缘

• paper: Innovations are disproportionately likely in the periphery of a scientific network [paper link]
• 中文阅读: 中文阅读

[yaoyz96]

多模态语言-视觉模型综述

• paper: A Survey on Multimodal Large Language Models [paper link]
• github: AwesomeMultimodal-Large-Language-Models
• 中文阅读: 中文阅读
• highlight
  • 大语言模型辅助视觉推理（CV-NLP）

[yaoyz96]

Nature子刊：针对临床疾病诊断的多模态表征学习模型

• paper: A transformer-based representation learning model with unified processing of multimodal input for clinical diagnostics [paper]
• github: IRENE
• 中文阅读: 中文阅读

[yaoyz96]

单张消费级显卡微调多模态大模型

• paper: Cheap and Quick: Efficient Vision-Language Instruction Tuning for Large Language Models [paper]
• github: LaVIN-lite
• project: https://luogen1996.github.io/lavin/
• 中文阅读: 中文阅读
• highlight
  • 在连续学习任务上的微调
  • 在小样本任务上的微调

[yaoyz96]

多模态大模型评测排行榜

• paper: MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models [paper]
• github: Awesome-Multimodal-Large-Language-Models
• 中文阅读: 中文阅读

[yaoyz96]

DegAE ：一种新的底层视觉预训练范式

• paper: DegAE: A New Pretraining Paradigm for Low-Level Vision [paper]
• 中文阅读: 中文阅读

[yaoyz96]

探索“AI 大统一理论”：科学启发的机器学习理论

• 中文阅读: 中文阅读
• highlight
  • grokking 现象（Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets）
  • 模型泛化性与 weight norm 的关系