xLSTM-Mixer: Multivariate Time Series Forecasting by Mixing via Scalar Memories First handles components of the time series linearly, up-projects, then refines by sLSTM layers, and then performs multi-view mixing. P: arxiv.org/abs/2410.16928

VLMs fail on #Bongard problems, a cognitive puzzle requiring abstract reasoning, limited data handling & contextual rules! #NeurIPS 2024 "System-2 Reasoning at Scale" Workshop PDF 👉 openreview.net/pdf?id=4Yv9tFH… A comparison to humans (soon on arXiv) shows a big gap

Our recent workshop paper puts Vision-Language Models like #GPT4o to the test with Bongard problems🧩 We show that VLMs struggle to identify concepts that are quite intuitive to humans. Still a long way to go for human-like visual reasoning! 🤖🧠 arXiv: arxiv.org/abs/2410.19546

Very nice overview of different xLSTM-based Methods

Had a blast collaborating on this project! If anyone’s interested and wants to chat at #NeurIPS, feel free to reach out to any of the authors—we’re all here in Vancouver!

This looks awesome!

Idea of having a rebuttal acknowledgement button in #ICML2025 is a big goof up. It reduces the accountability of the reviewers since we don't get anything even after answering everything. If it was a discussion then at least they are forced to say what they did or did not like!!

LAION proudly presents 2 state-of-the-art emotion detection models for voice and face, surpassing Gemini 2.5 Pro and Hume API. They are completely open under a CC BY 4.0 license, alongside a ~5,000-hour voice-acting dataset & 2 expert-annotated benchmarks. laion.ai/blog/do-they-s…

Who will be the first to annotate Emilia with this? :) Our new spripts for annotating speech along 40 emotion categories and 15 other dimensions like age, gender, arousal, softness, vulnerability, ... and generate pretty emotion captions: github.com/LAION-AI/emoti…

Today morning I asked the #neurips reviewers in my batch who just clicked the mandatory ack button to revisit and give reasons if or not their concerns were resolved by authors' rebuttals. Now we get an email from NeurIPS for the same. what is this NeurIPS Conference? 🤣😝

3 papers in #neurips25. 1 on formalizing meta-causal models, 1 on a novel optimization scheme for xLSTMs to realize Granger causal relationships in data and the last on a novel mixing scheme for xLSTMs proving to be SOTA on time series forecasting. Nice late birthday gift for me

Thrilled to announce our paper, "The Constitutional Filter: Bayesian Estimation of Compliant Agents" ☕️, has been accepted at IEEE IROS 2025! 🚀 In Hangzhou 🇨🇳 for IROS? Come see our talk and visit our poster! 🗓️ When: Tuesday, Oct. 21st 🗣️ Session: Sensor Fusion 2, 13:20-14:00