Markus Spanring (@MarkusSpanring) profile picture
Vienna, Austria
Joined January 2016
Markus Spanring retweeted
xLSTM for three-dimensional bin packing problem: arxiv.org/abs/2607.28257 xLSTM mixes item and candidate tokens with a recurrent action mixer (LRAM). xLSTM based OPAL has mean space utilization of 0.49 (15.1% improvement), while being fast. Promising for high-fidelity robotics.
1
19
151
9,960
Markus Spanring retweeted
# AI in Drug discovery just BROKE THROUGH a wall # A newer AI model, ConGLUDe, as fast but much more accurate than DrugCLIP. Instead on just 40K structure-based data, ConGLUDe is trained on 100M datapoints from ligand-based data P: arxiv.org/abs/2601.09693
BIG BREAKTHOUGH: A new AI tool could dramatically speed up the discovery of life saving medicines. Researchers at Tsinghua University created a new system called DrugCLIP, that can screen drug molecules against human proteins at a speed that makes traditional methods look ancient. > DrugCLIP uses deep contrastive learning to turn both molecules and protein binding pockets into vectors and match them almost instantly. > It screened 500 million molecules across 10,000 human proteins, covering half of the entire human druggable proteome. > The system completed 10 trillion molecule protein evaluations in a single day, roughly 10 million times faster than classic docking simulations. > They used AlphaFold2 to generate protein structures and then refined binding pockets with a custom tool called GenPack. > The model even identified compounds for TRIP12, a protein linked to cancer and autism that has resisted traditional drug-targeting approaches. All data and models are open access, so labs worldwide can now speed up early stage drug discovery.
10
59
3
291
26,525
xLSTM for Toxic Comment Classification: arxiv.org/abs/2510.17018 “On Jigsaw Toxic Comment benchmark, xLSTM attains 96.0% accuracy and 0.88 macro-F1, outperforming BERT by 33% on threat and 28% on identity_hate categories, with 15× fewer parameters and <50 ms inference latency.”
4
23
2,041
Markus Spanring retweeted
🚀 Excited to share our new paper on scaling laws for xLSTMs vs. Transformers. Key result: xLSTM models Pareto-dominate Transformers in cross-entropy loss. - At fixed FLOP budgets → xLSTMs perform better - At fixed validation loss → xLSTMs need fewer FLOPs 🧵 Details in thread
14
40
13
231
85,123
We release AB-UPT, a novel method to scale neural surrogates to CFD meshes beyond 100 million of mesh cells. AB-UPT is extensively tested on the largest publicly available datasets. 📄 arxiv.org/abs/2502.09692 🤗 huggingface.co/EmmiAI/AB-UPT 💻 github.com/Emmi-AI/AB-UPT
1
16
1
66
4,089
Parallelizable and state tracking and learnable information flow. Wowww. Super Work of Korbinian and team.
Ever wondered how linear RNNs like #mLSTM (#xLSTM) or #Mamba can be extended to multiple dimensions? Check out "pLSTM: parallelizable Linear Source Transition Mark networks". #pLSTM works on sequences, images, (directed acyclic) graphs. Paper link: arxiv.org/abs/2506.11997
17
1
113
5,613
Ever wondered how linear RNNs like #mLSTM (#xLSTM) or #Mamba can be extended to multiple dimensions? Check out "pLSTM: parallelizable Linear Source Transition Mark networks". #pLSTM works on sequences, images, (directed acyclic) graphs. Paper link: arxiv.org/abs/2506.11997
4
42
2
134
15,164
Markus Spanring retweeted
We’re excited to introduce TiRex — a pre-trained time series forecasting model based on an xLSTM architecture.
5
22
6
69
14,925
We are soooo proud. Our European-developed TiRex is leading the field—significantly ahead of U.S. competitors like Amazon, Datadog, Salesforce, and Google, as well as Chinese models from companies such as Alibaba.
We’re excited to introduce TiRex — a pre-trained time series forecasting model based on an xLSTM architecture.
1
12
2
74
3,991
Attention!! Our TiRex time series model, built on xLSTM, is topping all major international leaderboards. A European-developed model is leading the field—significantly ahead of U.S. competitors like Amazon, Datadog, Salesforce, and Google, as well as Chinese models from Alibaba.
We’re excited to introduce TiRex — a pre-trained time series forecasting model based on an xLSTM architecture.
2
24
109
5,815
Markus Spanring retweeted
Come by today at our posters in the Open Science for Foundation Models at 3pm (Hall4#5) #ICLR25 if you want to know more about Tiled Flash Linear Attention and xLSTM 7B!
11
42
3,003
Markus Spanring retweeted
🚀 Join us for an exclusive discussion on xLSTM 7B! To the future of fast and efficient LLMs w/ Maximillian Beck, PhD researcher at Johannes Kepler University & protégé of Mr. LSTM himself, Sepp Hochreiter. Hosted by @ceciletamura of @ploutosai app.ploutos.dev/streams/noct…
6
1
11
2,132
Markus Spanring retweeted
Amazing work with @KorbiPoeppel, @phillip_lippe, Richard Kurle, Patrick Blies, @gklambauer, Sebastian Böck and @HochreiterSepp (6/7)
1
3
10
1,354
Thrilled to announce two new developments at JKU and NXAI that are released today: - We scaled xLSTM to 7B parameters: linktr.ee/xlstm - For the people caring about state tracking capabilities, there's the new FlashRNN library: arxiv.org/abs/2412.07752
2
11
26
2,296
Markus Spanring retweeted
FlashRNN: Optimizing Traditional RNNs on Modern Hardware RNNs are more difficult to parallelize than Transformers, BUT.. This work performs hardware-optimization of RNNs in Triton and CUDA by optimizing kernels to the register level on modern GPUs. P: arxiv.org/abs/2412.07752
2
41
124
6,092
I am extremely glad to finally introduce our new parameter-efficient fine-tuning method called EVA (Explained Variance Adaptation). EVA extends LoRA with a data-driven initialization and adaptive rank allocation to speed up convergence and improve downstream performance. 1/n
5
74
5
435
66,434
Markus Spanring retweeted
Excited to present Vision-LSTM (ViL) at two ICML workshops (NGSM and LSFM). We also updated the arxiv paper: 🚀 Better training pipeline/models 🧮 Pre-training and inference FLOPS 📊 Evaluation on more datasets (segmentation, transfer classification) arxiv.org/abs/2406.04303
1
8
2
31
11,172
Update of the ArXiv Vision-LSTM (ViL) paper: - New and better models - FLOPS reported - More datasets (segmentation, transfer classifica.) arxiv.org/abs/2406.04303 Thanks Benedikt.
Excited to present Vision-LSTM (ViL) at two ICML workshops (NGSM and LSFM). We also updated the arxiv paper: 🚀 Better training pipeline/models 🧮 Pre-training and inference FLOPS 📊 Evaluation on more datasets (segmentation, transfer classification) arxiv.org/abs/2406.04303
2
19
96
7,883
Markus Spanring retweeted
Talking about #xlstm paper and a gripper - fantastic results by @HochreiterSepp @jo_brandstetter @gklambauer and team. Industrial AI application are waiting. Listen 🎧 on any catcher.
11
2
44
11,778
Markus Spanring retweeted
transformer killer just dropped xLSTM: Extended Long Short-Term Memory paper: arxiv.org/abs/2405.04517
10
68
11
638
91,358