Yayuan Li (李雅元)

Room 530

Dorm 21st, 163 Xianlin Street, Qixia Distirct

Nanjing City, Jiangsu province

Hi, I’m a Master’s Students at Nanjing University, under the supervision of Shi Yinghuan. To address the opportunities and challenges presented by large models (LMs), I mainly focus on Knowledge Transfer and Training-Free Adaptation. In particular, I am interested in model merging, aiming to efficiently utilize pre-trained models in the community and recombine them in a “LEGO-like” fashion. These techs seeks to decouple and enhance different capabilities of models by modular reassembly.

News

Dec 20, 2024	Our paper about CLIP-based FSL is accepted by AAAI25.
Jun 01, 2023	I was honored as an Outstanding Graduate of Beijing.
Dec 01, 2022	I have been awarded the National Scholarship.

Selected Publications

AAAI

Text and Image Are Mutually Beneficial: Enhancing Training-Free Few-Shot Classification with CLIP

Yayuan Li, Jintao Guo, Lei Qi, and 2 more authors

AAAI, Dec 2024

Abs DOI HTML PDF

Contrastive Language-Image Pretraining (CLIP) has been widely used in vision tasks. Notably, CLIP has demonstrated promising performance in few-shot learning (FSL). However, existing CLIP-based methods in training-free FSL (i.e., without the requirement of additional training) mainly learn different modalities independently, leading to two essential issues: 1) severe anomalous match in image modality; 2) varying quality of generated text prompts. To address these issues, we build a mutual guidance mechanism, that introduces an Image-Guided-Text (IGT) component to rectify varying quality of text prompts through image representations, and a Text-Guided-Image (TGI) component to mitigate the anomalous match of image modality through text representations. By integrating IGT and TGI, we adopt a perspective of Text-Image Mutual guidance Optimization, proposing TIMO. Extensive experiments show that TIMO significantly outperforms the state-of-the-art (SOTA) training-free method. Additionally, by exploring the extent of mutual guidance, we propose an enhanced variant, TIMO-S, which even surpasses the best training-required methods by 0.33% with approximately ×100 less time cost.