Sabitlenmiş Tweet
We are committed to making meaningful progress in machine learning research through open collaboration. Follow this 🧵to stay on top of our research contributions.
English
Cohere Labs
5.4K posts
Cohere Labs
@Cohere_Labs
@Cohere's research lab and open science initiative that seeks to solve complex machine learning problems. Join us in exploring the unknown, together.
Earth Katılım Nisan 2019
264 Takip Edilen26.6K Takipçiler
🐦 BIRDS: luma.com/CLBIRDS
👀 Computer Vision: luma.com/CLComputerVisi…
🦾 Embodied AI: luma.com/CLEmbodiedAI
🔢 ML Math: luma.com/CLMLMath
🤖 ML Agents: luma.com/CLMLAgents
🏥 ML Healthcare: luma.com/CLMLHealthcare
💼 ML Industry: luma.com/CLMLIndustry
🏫 EdTech: luma.com/CLEdTech
🔐 Privacy, Security & Policy: luma.com/CLPrivacySecur…
🧠 ML Understanding: luma.com/CLMLUnderstand…
🖼️ Multimodal: luma.com/CLMultimodal
💬 NLP: luma.com/CLNLP
🌍 Regional Africa: luma.com/CLRegionalAfri…
🌏 Regional Asia: luma.com/CLRegionalAsia
🤝 Research Connections: luma.com/CLResearchConn…
🏗️ Reinforcement Learning: luma.com/CLReinforcemen…
🦺 Safety and Alignment: luma.com/CLSafetyAlignm…
⚙️ Systems and Theory: luma.com/CLSystemsTheory
English
Did you know our open science community hosts 30+ events each month across a variety of ML Subfields? 🤩
Subscribe to dedicated event calendars for your favorite research areas and never miss a reading group, tech talk, workshop, or collaboration opportunity.
English
New blog from our community! PhD Fellow at Aarhus University, Ankita Maity, shares the story of her transition from industry to academia while collaborating on cultural AI research through open science initiatives.
🌍 Read the full post at labscommunity.cohere.com/blog/2026/a-co…
English
@UvA_Amsterdam @TheyCallMeMr_ @johnamqdang @mgalle @mziizm @ahmetustun89 @beyzaermis Announcing new research on cultural representation across LLM training data. 🌪️🏷️🕵️🗺️ Led by @ananya__sahu, @mhrnz_m, @mrdanieldsouza, @rsrchvslstrytlr, Julia Kreutzer, and @mziizm
📜Read the paper: arxiv.org/abs/2606.13808
English
@UvA_Amsterdam Announcing new research on Self-Verified RL using Soft-Rewards 📋, led by @TheyCallMeMr_, Pierre Clavier, @johnamqdang, @mgalle, @mziizm, @ahmetustun89, and @beyzaermis.
📜Read the paper: arxiv.org/abs/2605.28561
English
We are committed to making meaningful progress in machine learning research through open collaboration. Follow this 🧵to stay on top of our research contributions.
English
Cohere Labs retweetledi
We often treat multilinguality as a proxy for cultural competence. But speaking many languages doesn't mean understanding the culturally grounded meanings.
In the first paper of a larger bet we're making this year we ask: where does culture get lost in the LLM training pipeline?
Cohere Labs@Cohere_Labs
Announcing new research on cultural representation across LLM training data. 🌪️🏷️🕵️🗺️ Led by @ananya__sahu, @mhrnz_m, @mrdanieldsouza, @rsrchvslstrytlr, Julia Kreutzer, and @mziizm 📜Read the paper: arxiv.org/abs/2606.13808
English
Cohere Labs retweetledi
Come join us tomorrow for a presentation by the team (Haoquan + Jiafei) on MolmoAct2! Open to the public -> see details -> luma.com/2sstfy2b
@hq_fang @DJiafei
English
Announcing new research on cultural representation across LLM training data. 🌪️🏷️🕵️🗺️ Led by @ananya__sahu, @mhrnz_m, @mrdanieldsouza, @rsrchvslstrytlr, Julia Kreutzer, and @mziizm
📜Read the paper: arxiv.org/abs/2606.13808
English
Looking forward to this talk from @avicgoldfarb, hosted by @zanmuny, tomorrow, June 15 at 12pm ET. RSVP to join - luma.com/u8f79v4x
Cohere Labs@Cohere_Labs
On June 16, we'll be honoured to host @avicgoldfarb for a talk, "O-Ring Automation and the Economics of Bicycles for the Mind". Hosted by Research Fellow, @zanmuny, this event is part of our new series to learn from the experts whose work is influencing our own research. RSVP to join: luma.com/u8f79v4x
English
Cohere Labs retweetledi
Most cultural alignment approaches focus on inference-time interventions and assume models already contain sufficient cultural knowledge. We argue modern LLM pipelines suffer from a cultural data funnel, requiring a closer look at how culture is expressed in training data itself.
English
Cohere Labs retweetledi
Come join us on Monday, June 22 on my talk hosted by @Cohere_Labs on our recent “Attention Drift” paper.
If you are into transformer architecture, speculative decoding, or just curious about how “LLMs” work, this talk is for you.
harsha@Sree_Harsha_N
I'm really excited to host @dogacel0 at the ML Systems and Theory group at @Cohere_Labs community. Looking forward to learning about this very interesting phenomenon in speculative decoding and the insights it unlocks. When: 22 June at 16:00 GMT, Luma even in replies.
English
Cohere Labs retweetledi
I'm really excited to host @dogacel0 at the ML Systems and Theory group at @Cohere_Labs community. Looking forward to learning about this very interesting phenomenon in speculative decoding and the insights it unlocks. When: 22 June at 16:00 GMT, Luma even in replies.
English
Built as part of our recent Expedition Tiny Aya, with mentorship from @viraataryabumi
labscommunity.cohere.com/blog/2026/talk…
English
New blog from our community! @aktasbatuhann, @Swan_aes, and @promptlyconfuse introduce us to the Kids Multilingual Benchmark, a 2,312-prompt, 23-language benchmark for child–AI conversations that evaluates 4 production models & validates the LLM-as-judge pipeline with 5 independent judges. 🚸
English
Join us on June 26 as we welcome @MarineCarpuat to present "From Machine Translation to AI-Mediated Communication."
Cohere Labs' Sr Scientist, Julia Kreutzer, will host this instance of our series highlighting the work of researchers whose explorations are most impacting our current pursuits.
RSVP: luma.com/g034coqc
English
Our ML Math group is excited to host Debjyoti Paul for a session focused on "Softmax, Backprop, and the Autograd: The Hidden Machinery Behind Deep Learning" next week! 🎉
Learn more and join us for this insightful session: luma.com/3xmte81u
Thanks to community leads @KatrinaL2899 and @neel_ghoshal for organizing this event!
English
Huge thanks to @Rafi Ib Sultan for his presentation on "WalkGPT - Grounded Vision-Language for Pedestrian Navigation" 👏
Ready to learn how segmentation masks and advanced encoders are reshaping visual input handling with MSQP? Watch the recording now: youtu.be/od8AjajF8TQ
Special shoutout to community leads @cataluna84 and @Arkhymadhe for hosting this event!
YouTube
English
Cohere Labs retweetledi
One paper from three years ago has been influencing policymakers around the world about labour decisions regarding AI.
What happens when the gap between the evidence and new policy widens? Our report from @Cohere_Labs looks at these effects, as part of a new research direction exploring how AI impacts the future of work:
cohere.com/blog/the-futur…
English
Cohere Labs retweetledi
Great work from the Labs team on AI impact on Labour and ideas on how to better track the impact to workers.
Cohere Labs@Cohere_Labs
A single exposure score is driving the future of work debate. Our latest report examines what a fuller evidence base looks like, and what it would take to build it. See what is being missed. cohere.com/blog/the-futur…
English
Cohere Labs retweetledi
We mapped how occupational tasks have shifted over the last 3 years with and we want to know if it reflects your reality.
Check your occupation and tell us how AI has affected your work:
cohere.com/labormap
Marzieh Fadaee@mziizm
⚡️Excited to share our first report on the future(s) of work. We build models so we have a good view on what AI can and can't do, but it doesn't always transfer to how this technology is impacting every job. That's the gap we're trying to close in this work 🔦
English