Author

Shirley Li

@lixue421

articles: 10
total claps: 946
top claps: 178
avg read: 10.2 min
paywalled: 1
active: 2025-02-10 → 2026-05-17

Wiki topics

AI · AI · General · 9 EDU · Education & Learning · 9 ML · Machine Learning · 9 LLM · Large Language Models · 7 RAG · RAG & Retrieval · 2 🥊 · Combat Sports · 1 🏛️ · Architecture · 1

Top tags

#artificial-intelligence ·9 #deep-learning ·8 #large-language-models ·7 #deepseek ·6 #transformers ·3 #nlp ·3 #llm ·1 #deepseek-r1 ·1 #naturallanguageprocessing ·1 #matformer ·1

AI HUM

SH Shirley Li · AI Advances

DeepSeek-V3 Explained 3: Auxiliary-Loss-Free Load Balancing

How DeepSeek breaks the hidden bottleneck in MoE

Feb 10, 2025 · 10 min read · 178 claps #deepseek #large-language-models #mixture-of-experts

DeepSeek-V3 Explained 3: Auxiliary-Loss-Free Load Balancing

AI HUM

SH Shirley Li · Data Science Collective

DeepSeek-R1: Advancing LLM Reasoning with Reinforcement Learning

This is the seventh article in our DeepSeek series [1], where we will break down how DeepSeek-R1 is trained by exploring large-scale…

Apr 01, 2025 · 13 min read · 169 claps #deepseek #deepseek-r1 #large-language-models

DeepSeek-R1: Advancing LLM Reasoning with Reinforcement Learning

SPT HUM AI

SH Shirley Li

MatFormer: Train Once and Deploy Many

The secret weapon of Gemma 3n to let one model fit all.

Oct 22, 2025 · 16 min read · 164 claps #matformer #transformers #large-language-models

AI HUM

SH Shirley Li · Data Science Collective

DeepSeek Explained 6: All you need to know about Reinforcement Learning in LLM training

This is the sixth article in our DeepSeek series, where we will dive deeper into one of the key innovations in training strategies of…

Mar 18, 2025 · 15 min read · 148 claps #deepseek #reinforcement-learning #artificial-intelligence

DeepSeek Explained 6: All you need to know about Reinforcement Learning in LLM training

AI HUM

SH Shirley Li · Data Science Collective

DeepSeek Explained 4: Multi-Token Prediction

How DeepSeek achieves better balance between efficiency and quality in text generation

Feb 20, 2025 · 10 min read · 88 claps #deepseek #large-language-models #artificial-intelligence

DeepSeek Explained 4: Multi-Token Prediction

AI HUM

SH Shirley Li

DeepSeek Explained 8: Post-Training of DeepSeek-V3

This is the last article of our DeepSeek series ([1], [2]), finally we will cover the post-training techniques in DeepSeek-V3 [2].

Apr 07, 2025 · 5 min read · 55 claps #deepseek #artificial-intelligence #deep-learning

DeepSeek Explained 8: Post-Training of DeepSeek-V3

AI HUM

SH Shirley Li · Data Science Collective

DeepSeek Explained 5: DeepSeek-V3-Base

Innovations in pre-training strategies of DeepSeek-V3.

Mar 10, 2025 · 11 min read · 50 claps #deepseek #artificial-intelligence #large-language-models

SH Shirley Li

Transforming the Transformer (Part 1): The Rise and Fall of Efficient Attention

A deep dive into the rise and fall of X-formers, and why modern LLMs abandoned most of them.

May 17, 2026 · 7 min read · 40 claps paywalled #artificial-intelligence #llm #transformers

Transforming the Transformer (Part 1): The Rise and Fall of Efficient Attention

AI ART HUM

SH Shirley Li

Paper Explained 3: E5

How a simple architecture is transformed into a SOTA embedding model

Feb 18, 2025 · 10 min read · 32 claps #deep-learning #artificial-intelligence #e5

AI HUM

SH Shirley Li

Paper Explained 4: NV-Embed

How Nvidia turns decoders into efficient embedding models

Mar 03, 2025 · 6 min read · 22 claps #deep-learning #artificial-intelligence #transformers