Kelly Marchisio (St. Denis) Profile
Kelly Marchisio (St. Denis)

@cheeesio

Followers
2K
Following
3K
Statuses
640

Multilingualilty Lead @cohere. Formerly: PhD @jhuclsp, Alexa Fellow @amazon, dev @Google, MPhil @cambridgenlp, EdM @hgse 🔑🔑¬🧀 (@kelvenmar20)

Connecticut, USA
Joined June 2019
Don't wanna be here? Send us removal request.
@cheeesio
Kelly Marchisio (St. Denis)
24 days
RT @CohereForAI: On @scale_AI's private multilingual protocol, Aya Expanse is indexed as the best open-weights model In some languages we'…
0
19
0
@cheeesio
Kelly Marchisio (St. Denis)
26 days
@seb_ruder @AIatMeta You will be missed, @seb_ruder !
0
0
7
@cheeesio
Kelly Marchisio (St. Denis)
1 month
We worked hard to make Command R+ your best choice AI model for Arabic. Check it out!
@dani_avila7
Daniel San
1 month
At CodeGPT, we frequently receive inquiries from companies asking us to help them choose the best AI model for their specific needs. Recently, we worked with a client focused on Arabic software development. After thorough evaluation, Command R Plus from @cohere was the selected model! We conduct studies and guide companies to ensure they are using the best model on our platform If you're looking to integrate generative AI agents into your workflows, get in touch with us for consulting. Let’s collaborate to find the perfect solution for you.
0
1
21
@cheeesio
Kelly Marchisio (St. Denis)
1 month
@aidangomez Ok so it’s not just my feed. Especially in the last few weeks, it’s an absolute dumpster fire
0
0
1
@cheeesio
Kelly Marchisio (St. Denis)
2 months
RT @sarahookr: Enjoy Global-MMLU-lite. To deepdive into this evaluation set: This is cross-institutional work invo…
0
4
0
@cheeesio
Kelly Marchisio (St. Denis)
2 months
RT @PontiEdoardo: Is sparsity the key to conditional computation, interpretability, long context/generation, and more in foundation models?…
0
26
0
@cheeesio
Kelly Marchisio (St. Denis)
2 months
RT @p_nawrot: Do you want to learn how to build your own o1 - spend more compute on harder inputs, and less on easier ones? KV Cache takes…
0
3
0
@cheeesio
Kelly Marchisio (St. Denis)
2 months
C was first, but OCaml was the first I knew “well”. I still pass around functions like it’s nobody’s business, and will do nearly anything to avoid writing code with side-effects 🤢 (Over a decade later, though, I couldn’t write a line of OCaml to save my life 🙃)
@CogniCarbon
Carbon
2 months
Your first programming language shapes they way you solve problems. Really interesting read.
Tweet media one
0
0
8
@cheeesio
Kelly Marchisio (St. Denis)
2 months
RT @JustinTrudeau: Canada is a leader in AI because of companies like @Cohere. We are working with Cohere to build a cutting-edge AI data…
0
330
0
@cheeesio
Kelly Marchisio (St. Denis)
2 months
RT @JustinTrudeau: Le Canada est un leader en IA grâce aux entreprises comme @Cohere. On travaille avec Cohere pour mettre sur pied ici mê…
0
30
0
@cheeesio
Kelly Marchisio (St. Denis)
2 months
RT @CohereForAI: Aya Expanse: Combining Research Breakthroughs for a New Multilingual Frontier 🌿 Today, we release a technical report with…
0
31
0
@cheeesio
Kelly Marchisio (St. Denis)
2 months
We released Global-MMLU today! Check it out! 🌍
@CohereForAI
Cohere For AI
2 months
Is MMLU Western-centric? 🤔 As part of our cross-institutional work: 🥢 We conduct a large-scale cultural bias study on MMLU 🔍 Examine how cultural sensitivity impacts multilingual evaluations 🌍 Release Global-MMLU: a benchmark with MMLU translations in 42 languages
Tweet media one
1
3
23
@cheeesio
Kelly Marchisio (St. Denis)
3 months
@mayhewsw Her hyperparameters are well-tuned and for the current stage in training, her loss curve is among the lowest we’ve seen - and dropping fast. High hopes for this run!
0
0
3
@cheeesio
Kelly Marchisio (St. Denis)
3 months
@ecats_ Thank you, Cate! Baby says hellooooooooo and can’t wait to meet you!! 🤗
0
0
1
@cheeesio
Kelly Marchisio (St. Denis)
3 months
@johnamqdang Thanks, John!
0
0
1
@cheeesio
Kelly Marchisio (St. Denis)
3 months
@acyr_l Preliminary experiments tell us that a dramatically reduced step size and long training duration is best here - perhaps up to 4 years. We will, of course, evaluate earlier checkpoints in an effort to save energy via early-stopping
0
0
3
@cheeesio
Kelly Marchisio (St. Denis)
3 months
@AlexLoftus19 She’s got such a personality already 🤣
0
0
1
@cheeesio
Kelly Marchisio (St. Denis)
3 months
@bonadossou Thanks, Bona!
0
0
0
@cheeesio
Kelly Marchisio (St. Denis)
3 months
@arkadyark Thanks Arkady!
0
0
0
@cheeesio
Kelly Marchisio (St. Denis)
3 months
@soldni Thanks, Luca! Yes indeed, let the post training begin! 🤖
0
0
1