Alessandro Sordoni
@murefil
Followers
1K
Following
5K
Media
3
Statuses
556
Research Scientist / Manager, ML Team @MSFTResearch Montréal. Adjunct Professor at UdeM, @Mila_Quebec.
Montréal
Joined March 2010
We have few intern positions open in our ML team @ MSR Montreal, come work with @Cote_Marc @kim__minseon @LucasPCaccia @mathe_per @ericxyuan on reasoning, interactive envs/coding and LLM modularization. 🤯 @mathe_per and I will also be at #NeurIPS2024 so we can chat about this.
0
9
55
RT @ben_vandurme: I am growing an R&D team around Copilot Tuning, a newly announced effort that supports adaptation at a customer-specific….
0
14
0
RT @KateLobacheva: Our new paper got #ACL2025 oral! 🎉.If you're interested in LLM training dynamics, its phases, and how scaling affects th….
0
7
0
RT @MattVMacfarlane: Excited to attend #ICML2025 from Tue 15th to Sat 19th! Looking forward to connecting and discussing topics such as la….
0
1
0
RT @yoavartzi: @COLM_conf decisions are out, and so are we . The strength of submissions this year amazed us! Many many hard decisions 😩….
0
8
0
RT @hyunji_amy_lee: 🥳Excited to share that I’ll be joining @unccs as postdoc this fall. Looking forward to work with @mohitban47 & amazing….
0
27
0
Test-time scaling is all over the place right now, here we try to pack knowledge of particular documents in LoRAs (in "knowledge modules"), by performing expensive computation offline, such that test time computation is quick; i see this as precomputing and storing possible.
RAG and in-context learning are the go-to approaches for integrating new knowledge into LLMs, making inference very inefficient. We propose instead 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗠𝗼𝗱𝘂𝗹𝗲𝘀 : lightweight LoRA modules trained offline that can match RAG performance without the drawbacks
0
6
23
LLMs still struggle with text-based games such as Zork requiring memory, exploration, reasoning and rapid adaptation; great workshop to study these problems @ericxyuan @Cote_Marc.
0
3
18
RT @KushaSareen: 🚀 Introducing RL^V: unified RL & generative verifier that boosts MATH accuracy by 20%+ and delivers up to 32× more efficie….
0
46
0
RT @XinyiWang98: I’m attending #ICLR in Singapore! Also excited to share that I’m joining the Princeton Language and Intelligence Lab as a….
0
13
0
RT @rajammanabrolu: Introducing TALES - Text Adventure Learning Environment Suite. A benchmark of a few hundred text envs: science experime….
0
19
0
RT @Ar_Douillard: 30+ accepted papers. 6 oral papers. 6 guest speakers. join us at @iclr_conf on the 27th Hall 4 #3 for a full day of works….
0
30
0
RT @sarahookr: I am looking forward to seeing you all! . This is one of the key priorities for our lab and looking forward to supporting ev….
0
6
0
RT @tjingrant: Introducing Learned Asynchronous Decoding w/ friends from MIT/Google! LLM responses often have chunks of tokens that are sem….
0
15
0
RT @QVeraLiao: Another very cool work led by the very cool @sunniesuhyoung on how different components of LLM output influence appopriate r….
0
4
0
RT @Ar_Douillard: we received an outstanding interest for our ICLR workshop, and now we need your help! If you are a researcher interested….
docs.google.com
Please fill out this form if you are interested in being a reviewer for the ICLR 2025 Workshop on Modularity for Collaborative, Decentralized, and Continual Deep Learning (MCDC) The review period...
0
7
0
Rip David Lynch, thanks for gifting us with your work of art (especially Twin Peaks I m a huge fan of).
4
0
16
RT @arankomatsuzaki: SynthLabs + Stanford presents:. Towards System 2 Reasoning in LLMs: Learning How to Think With Meta Chain-of-Thought….
0
138
0
RT @a_kazemnejad: From OpenAI’s PPO, people start simplify it by removing its mechanisms, especially credit assignment, without performance….
0
33
0