🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Do you have opinions about the current state of reviewing at *CL conferences? Do you want to help? We (EMNLP 2025 PCs) want to hear from you: forms.office.com/r/P68uvwXYqf

Reinforcement Learning for Reasoning in Small LLMs: What Works and What Doesn't, arxiv.org/abs/2503.16219: The original DeepSeek-R1 paper showed that when developing small(er) reasoning models, distillation gives better results than pure RL. In this paper, researchers follow up on

The best way to learn #machinelearning is to: Avoid popular content that oversimplifies concepts, especially when written by individuals with no real ML experience. Dive into high-quality books and work through them diligently.

After a very successful 1st edition at European Conference on Computer Vision #ECCV2026 our Beyond Euclidean workshop is back, now at #ICCV2025 in October! We will update our webpage (sites.google.com/view/beyondeuc…) with information on our keynote speakers and open a call for full and short papers in the coming days!

The research reveals how Chatbot Arena rankings are systematically skewed to favor proprietary AI models through private testing, data access asymmetries, and biased deprecation policies.

What would truly open-source AI look like? Not just open weights, open code/data, but *open development*, where the entire research and development process is public *and* anyone can contribute. We built Marin, an open lab, to fulfill this vision:

🧠 Is curriculum learning a data-centric approach? 📊 No, by the strict definition (🔗 Forbes / Andrew Ng): It doesn’t modify data to improve its quality — e.g., labels, balance, coverage. 📈 Yes, in a broader sense (🔗 MIT DCAI 2023): It structures data — from easy to hard —

Happy to share that the #WiML2025 Symposium will be co-located with #ICML2025 in Vancouver! 🎉 🇨🇦 Many thanks to this year’s organizers: @mariehendriksen, Sophia Abraham, Meriem Mehri, Ninon Lizé Masclef, and Christianah Titilope Oyewale! Looking forward to meeting you there!

💎Friendly reminder: the original Bradley-Terry paper! jstor.org/stable/2334029 Just an appreciation post for those pursuing fundamental research. It's the slower path, demanding deep understanding, strong technical skills, and continuous grind. Often overshadowed by

It's already the case that people's free will gets hijacked by screens for hours a day, with lots of negative consequences. AI video can make this worse, since it's directly optimizable. AI video has positive uses, but most of it will be fast food for the mind.

China Cove, CA

✨ Just appeared in #informationgeometry Ke Sun and Frank Nielsen Sony CSL: "A geometric modeling of Occam’s razor in deep learning" #Deeplearning #Fisherinformation #HalfACenturyOfInformationGeometry Part 2 🔓 #OpenAccess link.springer.com/article/10.100…

🧵 Great to see reward-model calibration getting traction. two #ICLR2025 papers on the topic: 1️⃣ Post-hoc Reward Calibration : A Case Study on Length Bias. Zeyu Huang et al 2️⃣ Taming Overconfidence in LLMs: Reward Calibration in RLHF . Jixuan Leng et al 🔗

🚨 Calling all NLP researchers! WiNLP 2025 is seeking mentors for our pre-submission program 💫 Help early-career authors & underrepresented voices improve their work. 🗓️ Assignments: June 27 🗓️ Feedback due: July 11 Interested? Reply to join! #WiNLP #Mentorship #EMNLP2025

Authors & reviewers are usually the same people — but rebuttals + score updates overlap in the same week(s). 📆 What if we split the discussion period into two phases, minimizing cases where someone is both an author and reviewer in the same phase? 🔍 This gives us: – as

As NeurIPS review deadline is around the corner, please remember that you cannot use any non-local LLM like chatgpt/gemini for understanding the paper and drafting/revising your review as that breaks the confidentiality agreement.