Yelong Shen is a researcher in the field of artificial intelligence with a focus on machine learning and natural language processing. He has contributed to numerous scientific papers and is affiliated with Microsoft.
Shen is currently a Principal Investigator at the Vienna University of Technology. He previously held positions as a PhD Student at Meta AI & CentraleSupélec, where he researched self-supervised learning, and at Kent State University.
Shen has authored and co-authored several papers in the field of machine learning and natural language processing. His research interests include large language models, knowledge-grounded dialogue systems, text and query generation, and memory-efficient transformer architectures.
Some of his notable works include:
Efficient RLHF: Reducing the Memory Usage of PPO (2023) - This paper addresses the issue of high memory requirements for Reinforcement Learning with Human Feedback (RLHF) by proposing Hydra-RLHF, a method that integrates Supervised Fine-Tuning (SFT) and reward models to reduce memory usage and improve performance.
A Good Prompt Is Worth Millions of Parameters: Low-resource Prompt-based Learning for Vision-Language Models (2023) - This work focuses on prompt-based learning for vision-language models, proposing the FewVLM method to improve deployment and performance for real-world applications.
CAMERO: Consistency Regularized Ensemble of Perturbed Language Models with Weight Sharing (2023) - This paper introduces CAMERO, a consistency-regularized ensemble learning approach to improve model generalization while maintaining low memory costs by applying weight sharing and prediction consistency regularization.
Finding the Dominant Winning Ticket in Pre-Trained Language Models (2022) - The paper explores the Lottery Ticket Hypothesis in pre-trained language models, finding a "dominant winning ticket" that achieves competitive performance with a significantly reduced number of parameters.
Controllable Natural Language Generation with Contrastive Prefixes (2022) - This work proposes a lightweight framework for controllable natural language generation using contrastive prefixes, allowing for single-aspect and multi-aspect control while maintaining high linguistic quality.
UnitedQA: A Hybrid Approach for Open Domain Question Answering (2022) - This paper presents a hybrid approach for open-domain question answering, combining extractive and generative readers to leverage the strengths of both models and achieve improved performance.
Generation-Augmented Retrieval for Open-Domain Question Answering (2022) - The paper proposes Generation-Augmented Retrieval (GAR) for answering open-domain questions by augmenting queries with heuristically discovered relevant contexts, demonstrating improved performance compared to state-of-the-art retrieval methods.
MART: Memory-Augmented Recurrent Transformer for Coherent Video Paragraph Captioning (2021) - This work introduces the Memory-Augmented Recurrent Transformer (MART) architecture for generating coherent and relevant paragraph captions for videos by utilizing a memory module to summarize video segments and sentence history.
ReasoNet: Learning to Stop Reading in Machine Comprehension (publication date unknown) - This paper introduces ReasoNet, a neural network architecture for machine comprehension tasks that utilizes multiple turns to reason over the relationship between queries, documents, and answers.
Shen has also contributed to other notable works, including research on logical reasoning in LLMs, tool-integrated reasoning agents for mathematical problem-solving, and efficient vision-language pre-training.
Yelong Shen has frequently collaborated with researchers such as Weizhu Chen, Nan Duan, Yeyun Gong, Pengcheng He, and Minlie Huang, among others.
Youtube Title: soccer
Youtube Link: link
Youtube Channel Name: YeLonG Shen
Youtube Channel Link: https://www.youtube.com/@yelongshen3924
soccer
Youtube Title: What is Low-Rank Adaptation (LoRA) | explained by the inventor
Youtube Link: link
Youtube Channel Name: Edward Hu
Youtube Channel Link: https://www.youtube.com/@edwardjhu
What is Low-Rank Adaptation (LoRA) | explained by the inventor
Youtube Title: This Shaolin MASTER Almost KILLED Me
Youtube Link: link
Youtube Channel Name: Jesse Enkamp
Youtube Channel Link: https://www.youtube.com/@KARATEbyJesse
This Shaolin MASTER Almost KILLED Me
Youtube Title: LoRA: allow a high school student to train Large Language Model (GPT-3) with a gaming graphics card
Youtube Link: link
Youtube Channel Name: Olewave
Youtube Channel Link: https://www.youtube.com/@olewave
LoRA: allow a high school student to train Large Language Model (GPT-3) with a gaming graphics card
Youtube Title: E21: LoRA
Youtube Link: link
Youtube Channel Name: argmaxfm
Youtube Channel Link: https://www.youtube.com/@argmaxfm
E21: LoRA
Youtube Title: LoRA Explained
Youtube Link: link
Youtube Channel Name: Unify
Youtube Channel Link: https://www.youtube.com/@unifyai
LoRA Explained
Youtube Title: Cristiano Ronaldo
Youtube Link: link
Youtube Channel Name: YeLonG Shen
Youtube Channel Link: https://www.youtube.com/@yelongshen3924
Cristiano Ronaldo
Youtube Title: arxiv preprint - Language Models can be Logical Solvers
Youtube Link: link
Youtube Channel Name: AI Breakdown
Youtube Channel Link: https://www.youtube.com/@AGIBreakdown
arxiv preprint - Language Models can be Logical Solvers
Youtube Title: 《Ordinary Path》+ 《The brightest star in night sky》丨MangoTV
Youtube Link: link
Youtube Channel Name: 芒果TV音乐 MangoTV Music
Youtube Channel Link: https://www.youtube.com/@MangoTV-Music
《Ordinary Path》+ 《The brightest star in night sky》丨MangoTV
Youtube Title: THE PRIMARCHS - Sons of the Emperor | Warhammer 40k Lore
Youtube Link: link
Youtube Channel Name: The Remembrancer
Youtube Channel Link: https://www.youtube.com/@TheRemembrancer
THE PRIMARCHS - Sons of the Emperor | Warhammer 40k Lore
Youtube Title: Ronnie O'Sullivan vs Zhou Yuelong - Players Championship Snooker 2024 - First Round Live(Full Match)
Youtube Link: link
Youtube Channel Name: Snooker's Home
Youtube Channel Link: https://www.youtube.com/@SnookerHome-yw4ol
Ronnie O'Sullivan vs Zhou Yuelong - Players Championship Snooker 2024 - First Round Live(Full Match)
Youtube Title: 寶石Gem《野狼Disco》官方高畫質 Official HD MV | Disco Alaskan Wolves
Youtube Link: link
Youtube Channel Name: YOYOROCK 滾石移動
Youtube Channel Link: https://www.youtube.com/channel/UCCZasUiMCAk7rxq4HqZjmKQ
寶石Gem《野狼Disco》官方高畫質 Official HD MV | Disco Alaskan Wolves
Youtube Title: Finetuning LLMs with LORA
Youtube Link: link
Youtube Channel Name: Dr. Niraj Kumar (PhD, Computer Science)
Youtube Channel Link: https://www.youtube.com/@DrNirajRKumar
Finetuning LLMs with LORA
Youtube Title: Don't Mess With This Shaolin Monk
Youtube Link: link
Youtube Channel Name: BRUTAL TV
Youtube Channel Link: https://www.youtube.com/@Brutal_TV
Don't Mess With This Shaolin Monk
Youtube Title: CVPR18: Session 3-2C: Object Recognition & Scene Understanding V
Youtube Link: link
Youtube Channel Name: ComputerVisionFoundation Videos
Youtube Channel Link: https://www.youtube.com/@ComputerVisionFoundation
CVPR18: Session 3-2C: Object Recognition & Scene Understanding V
Youtube Title: EP1~90 Survival Game：My Cheats Can Be Stacked
Youtube Link: link
Youtube Channel Name: Light Novel
Youtube Channel Link: https://www.youtube.com/@LightNovel-EN
EP1~90 Survival Game：My Cheats Can Be Stacked
Youtube Title: What to expect at the 11th Beijing International Film Festival
Youtube Link: link
Youtube Channel Name: CGTN
Youtube Channel Link: https://www.youtube.com/@cgtn
What to expect at the 11th Beijing International Film Festival
Youtube Title: Machine Reading Comprehension
Youtube Link: link
Youtube Channel Name: Microsoft Developer
Youtube Channel Link: https://www.youtube.com/@MicrosoftDeveloper
Machine Reading Comprehension