Qiang Gao
@gaoqiang_nlp
a third year master student at Wuhan University, focusing on the natural language processing area. I'm actively looking for a PhD position for 2025Fall.
Since everyone wants to learn RL for language models now post DeepSeek, reminder that I've been working on this book quietly in the background for months. Policy gradient chapter is coming together. Plugging away at the book every day now. rlhfbook dot com
The new chatGPT 4o with Canvas system prompt:
This is the best use for of a drone I've ever seen 😭
don’t worry about it babe it’s just a brand
Congratulations to the new faculty members in @MITEngineering, including Media Lab alum Anna Huang (@huangcza)! Huang will hold a joint appointment in @MITEECS and MIT Music and Theater Arts. news.mit.edu/2024/school-en…
The Top ML Papers of the Week (May 27 - June 2): - SimPO - GNN-RAG - Attention as an RNN - Abacus Embeddings - Symbolic Chain-of-Thought - Contextual Position Encoding ...
I've been thinking about the many, MANY, DPO spinoff methods we've been seeing recently for rlhf. IPO, D2PO, CPO, ORPO, SPO, sDPO, KTO, DNO... Most claim they're "the best" but doesn't properly compare to related work. What do we do in alignment research? Thread 📚
🎉 Exciting News! 🎉 Just open-sourced my latest project: Llama3-based 8x8b-MoE model! 🚀 Extends llama3-8B-Instruct model with MoE architecture. Check it out & give it a star! github.com/cooper12121/ll…
If you really want to do something useful in AI, instead of training another tiny llama, pick up this project hazyresearch.stanford.edu/blog/2024-01-1… and train a 1B-parameter multilingual BERT with 32k input size. The code is here github.com/HazyResearch/m2. The data is all over @huggingface. The…
To close out 2023, here are 10 of the most interesting AI research advancements we shared on our feed this year — and where you can find more details on the work. 1️⃣ Segment Anything (SAM) A step toward the first foundation model for image segmentation. Details:…
Alibaba releases DreaMoving demo on Hugging Face A Human Video Generation Framework based on Diffusion Models demo: huggingface.co/spaces/jiayong…
Key vulnerabilities of GPT-4: 1. Fine-tuning API can remove or diminish safety guardrails, causing the model to produce harmful outputs or assist with dangerous requests 2. Fine-tuning can make the model generate targeted misinformation against public figures 3. Fine-tuning…
If you're not using AI, you're falling behind.
//#つぶやきProcessing int n=999999,p[]=new int[n],i,j; float t=1; void setup(){ size(800,800); for(i=2;i<n;i++)if(p[i]==0)for(j=i+i;j<n;j+=i)p[j]=i; } void draw(){ clear(); stroke(-1); for(i=2;i<n;i++)if(p[i]==0)circle(i*sin(i*t)/99+400,i*cos(i*t)/99+400,2); t+=1e-7; }
Foundation models are transforming society: in the past month alone, we've seen a flurry of releases! GPT-4, Claude, PaLM API, Alpaca, Dolly, Jurassic-2, PaLM-E, GPT4All, Cerebras-GPT, OpenFlamingo, ... We built Ecosystem Graphs to track their footprint: crfm.stanford.edu/ecosystem-grap…
4 essential books anyone should read: • Machine Learning with PyTorch and Scikit-Learn • Transformers for NLP • Deep Learning with Python • Designing Machine Learning Systems
United States Trends
- 1. Lakers 49.2K posts
- 2. Dillon Brooks 6,704 posts
- 3. Bron 24.2K posts
- 4. Giants 84.6K posts
- 5. #WWERaw 70.1K posts
- 6. Patriots 129K posts
- 7. Dart 35.9K posts
- 8. Suns 18.7K posts
- 9. Drake Maye 24.1K posts
- 10. Collin Gillespie 1,987 posts
- 11. #AvatarFireAndAsh 2,780 posts
- 12. Diaz 33.5K posts
- 13. Devin Williams 7,687 posts
- 14. Ryan Nembhard 4,057 posts
- 15. Pats 16.1K posts
- 16. Gunther 15K posts
- 17. STEAK 10.3K posts
- 18. James Cameron 4,326 posts
- 19. Devin Booker 2,593 posts
- 20. Joe Schoen 3,429 posts
Something went wrong.
Something went wrong.