Chulin Xie
@ChulinXie
Followers
667
Following
687
Media
6
Statuses
53
CS PhD student at UIUC and student researcher @GoogleAI ; Ex research intern @MSFTResearch @NvidiaAI
Joined March 2019
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
田中さん
• 539934 Tweets
田中敦子さん
• 516372 Tweets
声優さん
• 239296 Tweets
LINGORM THE STYLER
• 191846 Tweets
攻殻機動隊
• 169118 Tweets
Fate
• 157570 Tweets
ALRIGHT POSTER
• 142823 Tweets
ネルフェス
• 100357 Tweets
ショック
• 79798 Tweets
草薙素子
• 77125 Tweets
Divine Play Of God Kabir
• 67490 Tweets
#西園寺さんは家事をしない
• 61742 Tweets
フランメ
• 54882 Tweets
ラストマイル
• 44298 Tweets
メディアさん
• 42510 Tweets
초코우유
• 34861 Tweets
カーミラさん
• 30392 Tweets
Leao
• 29584 Tweets
OUR BRILLIANT RENJUN
• 29545 Tweets
ベヨネッタ
• 27789 Tweets
Zize
• 27714 Tweets
キャスター
• 27142 Tweets
Arhan
• 26522 Tweets
メアリー
• 19074 Tweets
岡田将生
• 15519 Tweets
レインドット
• 15438 Tweets
#INZM_10Mviews
• 15135 Tweets
スーパームーン
• 14341 Tweets
リサリサ先生
• 14118 Tweets
#يواكيم_اندرسن_مطلب_الاتحاديين
• 11981 Tweets
#それ言ってどうすんだ祭
• 11890 Tweets
美城常務
• 10489 Tweets
Pinned Tweet
Some text data is private & cannot be shared... Can we generate synthetic replicas with privacy guarantees?🤔
Instead of DP-SGD finetuning, use Aug-PE with inference APIs! Compatible with strong LLMs (GPT-3.5, Mistral), where DP-SGD is infeasible.
🔗 [1/n]
4
25
88
Excited to see the release of the book 🥳 and grateful for the opportunity to contribute a chapter. Big thanks to the three editors!
@pinyuchenTW
@LamMNguyen3
@nghiaht87
Happy to share the release of the book "Federated Learning: Theory and Practice" that I co-edited with
@LamMNguyen3
@nghiaht87
, covering fundamentals, emerging topics, and applications. Kudos to the amazing contributors to make this book happen!
@ElsevierNews
@sciencedirect
2
10
61
1
2
30
Our benchmark is on
@huggingface
leaderboard!!
Super excited to set up the LLM safety & trustworthiness leaderboard on Huggingface, and we will keep adding new safety perspectives. Here we evaluate (open & close) LLMs and compressed LLMs. Looking forward to more exciting evaluations to assess and enhance LLM safety!!! 🥳
1
8
51
0
4
27
How many adversarial users (instances) that a user-level (instance-level) DPFL mechanism can tolerate with certified robustness? 🧐
Check out our paper!
👥 Collaboration with
@long_yunhui
,
@pinyuchenTW
,
@Cubeeli
,
@sanmikoyejo
,
@uiuc_aisecure
🥳
1
3
10
Joint work with awesome collaborators
@lin_zinan
, Arturs Backurs,
@gopisivakanth
,
@DaYu85201802
,
@HuseyinAInan
,
@HarshaNori
, Haotian Jiang,
@HuishuaiZhang
,
@YinTatLee
,
Sergey Yekhanin at
@MSFTResearch
, and
@uiuc_aisecure
! 🥳
Code:
[7/n]
0
0
7
Our adaptive sequence length mechanism can enable GPT-3.5 to match the length of private data across iterations. [6/n]
2
0
4
4 steps: 1) first draw random samples from an LLM, and then iteratively improve them by 2) computing the nearest neighbor, 3) selecting (with DP) the most similar ones to the private dataset, and 4) querying the LLM to generate more of such related samples. [2/n]
1
0
3
Aug-PE is compatible with open-source LLMs, where DP finetuning is hard to implement due to the need to calculate the per-sample gradient in DP-SGD. [4/n]
1
0
2
Aug-PE (w/ inference only) can match DP finetuning with the same model under privacy budget epsilon = 4, 2, 1 in some cases, and outperform it with stronger closed-source LLMs like GPT-3.5, where DP finetuning is infeasible! [3/n]
1
0
3
Aug-PE (w/ inference APIs) is more efficient than DP-SGD finetuning on the Yelp dataset. The running time of Aug-PE mainly depends on the number of API calls (associated with the hyperparameter L). [5/n]
2
0
2
@PandaAshwinee
Thanks! We exactly used downstream objectives to evaluate syn data. For Yelp/OpenReview syn data, we evaluate accuracy on downstream classification tasks. For PubMed syn data, we evaluate next-word prediction accuracy of a downstream model, inspired by
2
0
1