Switzerland
Joined August 2015
- Tweets50
- Following79
- Followers127
- Likes61
Beatriz Borges retweeted
🗒️Can we meta-learn test-time learning to solve long-context reasoning?
Our latest work, PERK, learns to encode long contexts through gradient updates to a memory scratchpad at test time, achieving long-context reasoning robust to complexity and length extrapolation while scaling efficiently at inference.
PERK can be applied to existing pretrained language models without requiring architectural or parameter modifications to the base model.
#LLM #LongContext
Find out how PERK operates and performs 👇
Beatriz Borges retweeted
🚨New Preprint!!
Thrilled to share with you our latest work: “Mixture of Cognitive Reasoners”, a modular transformer architecture inspired by the brain’s functional networks: language, logic, social reasoning, and world knowledge.
1/ 🧵👇
Beatriz Borges retweeted
I am recruiting 2 PhD students for Fall'25 @csaudk to work on bleeding-edge topics in #NLProc #LLMs #AIAgents (e.g. LLM reasoning, knowledge-seeking agents, and more).
Details: cs.au.dk/~clan/openings
Deadline: May 1, 2025
Please boost!
cc: @WikiResearch @AiCentreDK @CPH_SODAS
Beatriz Borges retweeted
🚨 New Preprint!!
LLMs trained on next-word prediction (NWP) show high alignment with brain recordings. But what drives this alignment—linguistic structure or world knowledge? And how does this alignment evolve during training? Our new paper explores these questions. 👇🧵
Beatriz Borges retweeted
Are LLMs linguistically productive and systematic in morphologically-rich languages as good as humans?
No 🤨 Our new NAACL 2025 paper (arxiv.org/abs/2410.12656) reveals a significant performance gap between LLMs and humans in linguistic creativity and morphological generalization.
Beatriz Borges retweeted
🚨 New Paper!
Can neuroscience localizers uncover brain-like functional specializations in LLMs? 🧠🤖
Yes! We analyzed 18 LLMs and found units mirroring the brain's language, theory of mind, and multiple demand networks!
w/ @GretaTuckute, @ABosselut, & @martin_schrimpf
🧵👇
Beatriz Borges retweeted
🚀 Introducing INCLUDE 🌍: A multilingual LLM evaluation benchmark spanning 44 languages!
Contains *newly-collected* data, prioritizing *regional knowledge*.
Setting the stage for truly global AI evaluation.
Ready to see how your model measures up?
#AI #Multilingual #LLM #NLProc
Beatriz Borges retweeted
Latest generative AI models show impressive abilities, but where do they stand w.r.t creativity? Our new preprint "Creativity in AI: Progresses and Challenges" dives into recent advances and challenges in AI’s creativity, across problem-solving, linguistics, art, and science.
Beatriz Borges retweeted
We present LoRA-XS, an extremely parameter-efficient #LoRA variant, reducing the LoRA’s trainable parameters by more than 100x while performing competitively across different scales and benchmarks!
Find out how we did this ⬇️
Beatriz Borges retweeted
New year, new preprint 🎇🦾🧐!
We propose using structured negotiation games as dynamic benchmarks for language models (LMs).
LMs are being used to create *dynamic* agents but benchmarks have remained *static*. Structured negotiation games solve this!
arxiv.org/abs/2401.04536
Beatriz Borges retweeted
Do multilingual language models (MultiLMs) have what it takes to reason across languages?
Our #EMNLP2023 #NLProc paper proposes a new attention mechanism that considerably improves the cross-lingual generalization of MultiLMs!
Beatriz Borges retweeted
Ever wonder how AI writing support is biasing student responses? 📝✍️🤖 Our new #EMNLP2023 Findings paper unravels gender bias from LLMs downstream. Through 8 analyses with 200+ students, we present a case study where ✨gender bias did not transfer✨ from models to students! – at Lausanne, Switzerland
Beatriz Borges retweeted
Does your NLP system use commonsense when it matters? Hard to say when we use benchmarks that test models in artificial QA scenarios. We built CRoW🐦⬛, a new benchmark that evaluates model commonsense reasoning in real-world tasks where we actually use it
mete.is/crow🧵
Beatriz Borges retweeted
Absolutely thrilled that @silin_gao @obiwit @denizbayazit 's paper: "PeaCoK: Persona Commonsense Knowledge for Consistent and Engaging Narratives" received an Outstanding Paper Award at @aclmeeting #ACL2023NLP
This Post is from a suspended account. Learn more
Beatriz Borges retweeted
Come check out the poster for our #ACL2023 Outstanding Paper, PeaCoK: Persona Commonsense Knowledge for Consistent and Engaging Narratives, presented by @obiwit this morning at 11 AM in the Frontenac Ballroom!
Paper: arxiv.org/abs/2305.02364
Beatriz Borges retweeted
🚨NEW PAPER ALERT: We know that in-context learning (ICL) can be very sensitive to a host of design decisions, but did you know the task corpus itself is also a major source of bias?
Check out our recent ACL 2023 paper: arxiv.org/abs/2305.19148
1/
Beatriz Borges retweeted
📢New paper to appear at #acl2023nlp: arxiv.org/abs/2212.10534
Human-quality counterfactual data with no humans! Introduce DISCO, our novel distillation framework that automatically generates high-quality, diverse, and useful counterfactual data at scale using LLMs.
