@Cohere's research lab and open science initiative that seeks to solve complex machine learning problems. Join us in exploring the unknown, together.
Earth
- Tweets 5,619
- Following 273
- Followers 27,994
- Likes 3,674
Pinned 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.
32
60
391
Cohere Labs retweeted
‼️ Calling all AI for education enthusiasts! Excited to team up with @Cohere_Labs next Wednesday to discuss our multi-agent learning system at @ScholeAI. We'll also be discussing an RCT we ran in collaboration with @Berkeley_EECS and @EPFL_en called "When Should a Lesson Adapt, and for Whom? A Randomized Study to Understand When Adaptive Instruction Improves Learning" It'll foreshadow our recently conducted research, and the results are super interesting. You won't want to miss it, it's streaming online! See you there. :)
luma.com/qhf4qvv3?tk=80k3Jb
1
9
331
Cohere Labs retweeted
Reminder that this happening soon! Super excited for it. RL and Robotics friends, you won't want to miss this!
Ready for inspiration, discussion & community collabs? 💡🤝
Join us October 28th for research pitches from @tw_killian (Asst. Professor at BYU and Principal Scientist, AI Research at Lila Sciences) & Sha Yi (Research Scientist at Physical Intelligence, incoming NYU prof), whose work spans RL, LLM reasoning & robot learning.
Thanks to @TheEsraaSaleh, Ruchira Dhar and @Ujan55906689 for organizing this session! 🔥
RSVP to join: luma.com/igac551y
2
1
4
329
Cohere Labs retweeted
As a spin-off of CulturalRiddles, @Poindexter_Labs turned the riddles into cultural crosswords, and they’re actually pretty cool.
Here are the Canadian and Persian editions. Shout-out to @Jocelyn_K_DArcy and amazing her team!
#culturalriddles [2/4]
1
1
4
296
Open science is only possible because of the people who make it a collective effort. 💙
A huge thank you to the 160+ riddle writers who shared their time, knowledge, and cultures to help build CulturalRiddles. By contributing riddles in their own languages, they helped create a benchmark that challenges AI models to go beyond words and facts to better understand cultural context.
🌟 Thank you to all our contributors for making this work possible!
#CulturalRiddles
2
8
51
1,978
“Knowing a language without its culture is like reading a restaurant menu without ever tasting the food: you can name every ingredient, yet miss what the meal means to the people eating it.”
In this blog, Vinod Anbalagan, Tianyi Hu, Madeline Smith, Julia Kreutzer and some of the research leads on CulturalRiddles, share the story and key findings behind our benchmark exploring what AI models miss when it comes to cultural understanding. cohere.com/research/papers/c…
(Key takeaways below 🧵)
3
4
16
1,345
Models also struggled to recognize what they didn’t know. They frequently fabricated answers rather than expressing uncertainty, showing how fluent responses can mask gaps in cultural understanding.
1
4
198
CulturalRiddles highlights why multilingual AI evaluation needs to go beyond language coverage. Cultural knowledge, context, and lived experience need to be a part of the story. 🧩
Learn more: cohere.com/research/papers/c…
#CulturalRiddles
4
191
We are committed to making meaningful progress in machine learning research through open collaboration. Follow this 🧵to stay on top of our research contributions.
32
60
391
Announcing new research on culturally grounded riddles as a benchmark for evaluating LLM cultural understanding across 62 cultures. 🧩
Work done in our open science community including 150+ authors across 45 countries.
Learn more: cohere.com/research/papers/c…
1
144
Announcing new research on heterogeneous multi-agent teams, demonstrating that strategic compute allocation—cheap planners for planning, stronger models for implementation—achieves near-optimal rewards at a fraction of the cost of homogeneous teams.
Work led by @keenansamway, @mrdanieldsouza, Omar Darwiche Domingues, @mziizm, Tan Yi-Chern, @nouhadziri
📜: cohere.com/research/papers/s…
1
1
141
Announcing new research on the Agentic Task Ecosystem (ATE), a tagged dataset revealing developer tool usage patterns across occupations via MCP tools.🔍
Work led by @zanmuny, Campbell Lund, @rsrchvslstrytlr, @AidanPeppin, @mziizm
📜: cohere.com/research/papers/t…
39
Announcing our research on data mixing as the pillar of generalization for in-language reasoning. 🔀
Work led by @mhrnz_m, Ananya Sahu, @alexrs95, @mrdanieldsouza, Alexandre Bérard, @rsrchvslstrytlr, @mziizm, Julia Kreutzer
Read the paper: cohere.com/research/papers/b…
1
3
143
Cohere Labs retweeted
Language models claim to know many languages, but do they understand many cultures? 🤔
Very honored to contribute to newly released CulturalRiddles benchmark from @Cohere_Labs 🌐
#culturalriddles
3
2
25
672
Cohere Labs retweeted
AI chatbots struggle to understand cultural context, a new study finds trib.al/8o8XVnM
1
2
2,540
Cohere Labs retweeted
🧵 Proud to be a co-first author on CulturalRiddles: 6,191 riddles written by native speakers across 62 cultures.
The best model scores ~79% overall. In some communities, even frontier models top out at 22%.
We're still looking for contributors: github.com/Cohere-Labs-Commu… [1/9]
Can you solve a riddle rooted in another culture? 🧩
Our latest research, CulturalRiddles, explores whether AI models can understand the cultural context behind human-written riddles from around the world.
Test your knowledge with our quiz! culturalriddles.ai/
1
5
15
570
Cohere Labs retweeted
Can AI solve traditional Luganda & Lumasaba riddles? 🇺🇬"
Nze nnyumba ya bbugo ey'abaana. Ndi ani?"
Cultural reasoning tests common sense in ways standard benchmarks miss.
Try testing yourself (and the models) on the interactive quiz:👉 culturalriddles.ai/
#culturalriddles
1
5
228
LLMs are getting better and better at math & science. But surprisingly, many can't solve a basic riddle.
@Cohere_Labs' CulturalRiddles is a new benchmark measuring culture as intertwined, lived experiences. True to open science, it was made by an unlikely crew.
9
9
45
4,021
Cohere Labs retweeted
Fluent ≠ Culturally Aware. 💡
Delighted to collaborate with the Cohere Labs Community on our latest research!
We found that even top models struggle with local cultural nuances, often guessing or hallucinating when hit with regional context.
Check out: cohere.com/research/papers/c…
Can you solve a riddle rooted in another culture? 🧩
Our latest research, CulturalRiddles, explores whether AI models can understand the cultural context behind human-written riddles from around the world.
Test your knowledge with our quiz! culturalriddles.ai/
1
4
327