Selective Reflection-Tuning: Student-Selected Data Recycling for LLM Instruction-Tuning
Paper • 2402.10110 • Published • 3
Shwai He
shwai-he
AI & ML interests
Deep Learning, Mechine Learning, Natural Language Processing.
Recent Activity
authored a paper 1 day ago
Selective Reflection-Tuning: Student-Selected Data Recycling for LLM
Instruction-Tuning authored a paper 1 day ago
What Matters in Transformers? Not All Attention is Needed authored a paper 1 day ago
Capacity-Aware Inference: Mitigating the Straggler Effect in Mixture of
ExpertsOrganizations
authored 11
papers 1 day ago
What Matters in Transformers? Not All Attention is Needed
Paper • 2406.15786 • Published • 33
Capacity-Aware Inference: Mitigating the Straggler Effect in Mixture of Experts
Paper • 2503.05066 • Published • 4
SymRTLO: Enhancing RTL Code Optimization with LLMs and Neuron-Inspired Symbolic Reasoning
Paper • 2504.10369 • Published • 2
CogniPair: From LLM Chatbots to Conscious AI Agents -- GNWT-Based Multi-Agent Digital Twins for Social Pairing -- Dating & Hiring Applications
Paper • 2506.03543 • Published • 1
CoIn: Counting the Invisible Reasoning Tokens in Commercial Opaque LLM APIs
Paper • 2505.13778 • Published • 5
Dense Video Understanding with Gated Residual Tokenization
Paper • 2509.14199 • Published • 3
Understanding and Harnessing Sparsity in Unified Multimodal Models
Paper • 2512.02351 • Published • 4
Making Large Language Models Efficient Dense Retrievers
Paper • 2512.20612 • Published • 4
ThinkJEPA: Empowering Latent World Models with Large Vision-Language Reasoning Model
Paper • 2603.22281 • Published • 17
Demystifying When Pruning Works via Representation Hierarchies
Paper • 2603.24652 • Published • 19
submitted a
paper to Daily Papers 7 days ago
authored 8
papers over 1 year ago
Router-Tuning: A Simple and Effective Approach for Enabling Dynamic-Depth in Transformers
Paper • 2410.13184 • Published • 3
Reflection-Tuning: Data Recycling Improves LLM Instruction-Tuning
Paper • 2310.11716 • Published • 6
Merging Experts into One: Improving Computational Efficiency of Mixture of Experts
Paper • 2310.09832 • Published • 1
Vega-MT: The JD Explore Academy Translation System for WMT22
Paper • 2209.09444 • Published
MerA: Merging Pretrained Adapters For Few-Shot Learning
Paper • 2308.15982 • Published
Superfiltering: Weak-to-Strong Data Filtering for Fast Instruction-Tuning
Paper • 2402.00530 • Published • 2
Reformatted Alignment
Paper • 2402.12219 • Published • 17
Loki: Low-Rank Keys for Efficient Sparse Attention
Paper • 2406.02542 • Published