🫗 An LLM's "private" reasoning may leak your sensitive data! 🎉 Excited to share our paper "Leaky Thoughts: Large Reasoning Models Are Not Private Thinkers" was accepted at #EMNLP main! 1/2

🔎Does Conversational SEO actually work? Our new benchmark has an answer! Excited to announce our new paper: C-SEO Bench: Does Conversational SEO Work? 🌐 RTAI: researchtrend.ai/papers/2506.... 📄 Paper: arxiv.org/abs/2506.11097 💻 Code: github.com/parameterlab... 📊 Data: huggingface.co/datasets/par...

📅 March 28th is an exciting day for scientific networking! 🚀 Join our Connect sessions on VLM, FedML, and LLMAG—top researchers will present their latest findings! 🔗 Zoom Link: us06web.zoom.us/j/8409727334... #ResearchTrend #AI #Networking #VLM #FedML #LLMAG

ResearchTrend.AI is hiring a data engineer - python, airflow, and postgre talents ate required (+experience with LLMs would be awesome). Please apply to [email protected] or dm me!!!

Today's featured community is a unique one: Optimal Transport. It sits at the intersection of several machine learning subfields. Follow the community to receive updates: researchtrend.ai/communities/OT

Audio LLM community is now available on ResearchTrend.AI! researchtrend.ai/communities/... Check out our cool video :)

AI in education has gained significant momentum since the beginning of the "LLM era" (2023). Check out the papers at researchtrend.ai/communities/....

🚀 We launched the Graph Neural Networks (GNN) community on ResearchTrend.AI! Explore papers: researchtrend.ai/communities/... GNNs extend neural networks to graph data. The field boomed (2017-2020), plateaued (2020-21), and is now declining. Is this success, hype, or limits? 🤔 Share your thoughts!

Rejected at #ICLR2025: Scalable Ensemble Diversification for OOD Generalisation and Detection - Current approach: Ensemble diversification relies on a separate source of OOD samples for training models to diversify outputs. - Ours: Use samples from the training set itself to diversify ensembles.

Another #ICLR2025 paper: Intermediate Layer Classifiers for OOD generalization - Common approach: Use penultimate-layer features of pre-trained models for downstream tasks. - Our recommendation: Explore lower-layer features. You'll likely find better layers for OOD generalisation.

Finally accepted! #ICLR2025 Do Deep Neural Network Solutions Form a Star Domain? - Known: as DNN width --> infty, solutions become convex modulo permutations of neurons. - What's new: as DNN width --> infty, solutions **first form a star domain** and then a convex domain modulo permutations.

🎉 Happy New Year! We’re excited to announce the launch of our email digest feature on ResearchTrend.AI! 🚀 You can now receive daily updates from the research communities you follow. Stay informed! 1/3

Here’s a podcast I did with Danish journalist Lone Frank during Folkemødet on Bornholm last summer. The opening blurb is in Danish, and the rest is in English. We talk about my move from the US to DK, and the Pioneer Centre for AI’s unique approach to research.

🚀 Exciting subdomains of language models with immense potential and growing attention: 1/5

🎄 This holiday season, we have a special present for all AI-yites! 🎄 We’re announcing 11 new research communities at researchtrend.ai (@researchtrend.ai) to help you stay updated. We group them under 4 larger areas: 🧵

Test results confirm I had influenza B these past few days. It was rough! I suspect NeurIPS was the source. It seems like many fell ill after the event. 😅

What if we viewed ML generalisation not just as a machine's ability to perform well on unseen data, but as a form of human-to-ML communication? 1/8

Flying back to Seoul after an incredible time at NeurIPS in Vancouver! I’ll be in Seoul until 2 January before returning to Germany.

Curious about why embedding-space ensembles are a compelling alternative and how to make them work? 🤔 Check out our work on Sat. Dec 14 @neuripsconf.bsky.social Workshops! SSL: * 12:30 pm - 1:50 pm * West Meeting Room 202-204 @unireps.bsky.social: * 3:45 pm - 5:00 pm * West Exhibition Hall C, B3

Giving a talk tomorrow at #NeurIPS2024 on the exciting topic of explainability!

Excited for #NeurIPS2024! Here are some highlights. - Spotlight paper by @bmucsanyi.bsky.social on uncertainty disentanglement (epistemic vs aleatoric): bsky.app/profile/bmuc... - ATTRIB workshop by @elinguyen.bsky.social: bsky.app/profile/elin... 1/2

Last Friday, we hosted the first ResearchTrend Connect session! It was reminiscent of the nice discussions from the 2010s that brought together experts in the same domain (which we largely miss now).

This is happening now! researchtrend.ai/social-event...

Excited to host the first ResearchTrend.AI Connect sessions this Friday! 🎉 🎙️ Topics - Membership Inference Attack on Language Models: researchtrend.ai/social-event... - Privacy Issues in Language Models: researchtrend.ai/social-event... More info below 👇

Decomposing aleatoric and epistemic uncertainty is harder than you think. Meet us at #NeurIPS2024 for more.

Anyone coming to #NeurIPS? Let's meet :)

There's an internship opening at @parameterlab.bsky.social : parameterlab.de/careers The research outputs have been quite successful so far: researchtrend.ai/organization...

From tomorrow, the Bluesky bot @researchtrend.ai will provide a daily digest of the latest papers in each AI community. More info: researchtrend.ai

NeurIPS Conference is now Live on Bluesky! -NeurIPS2024 Communication Chairs

For social media, growth becomes a reason for even greater growth.

Despite a massive head start, BlueSky has now overtaken Threads in the US 👇

🚨📄 Exciting new research! Discover when and at what scale we can detect if specific data was used in training LLMs — a method known as Membership Inference (MIA)! Our findings open new doors for using MIA as potential legal evidence in AI. 🧵 https://arxiv.org/abs/2411.00154

Check out researchtrend.ai and get quick updates on the AI communities you follow.

New here? Interested in AI/ML? Check out these great starter packs! AI: go.bsky.app/SipA7it RL: go.bsky.app/3WPHcHg Women in AI: go.bsky.app/LaGDpqg NLP: go.bsky.app/SngwGeS AI and news: go.bsky.app/5sFqVNS You can also search all starter packs here: blueskydirectory.com/starter-pack...

This is an instance of positive re-purposing of technology that was built for LLM jailbreaking. I'm also quite proud of the collaborative work here.

me watching Bluesky's user count grow by another million in a day hello and welcome to all 19M of you! 🥳