Leon Klein
@leonklein26
Followers
440
Following
2K
Media
13
Statuses
133
PhD student @FU_Berlin @bifoldberlin, working on enhancing MD with ML. Former Visiting Researcher @MSFTResearch and ML Intern @DEShawResearch.
Joined January 2021
Excited to share our latest preprint: "Transferable Boltzmann Generators"! We propose a framework based on flow matching and demonstrate transferability on dipeptides. Work done with amazing @FrankNoeBerlin. Check it out here:
10
17
127
I'm excited to share FlowMol3! the 3rd (and final) version of our flow matching model for 3D de novo, small-molecule generation. FlowMol3 achieves state of the art performance over a broad range of evaluations while having โ10x fewer parameters than comparable models.
1
3
16
After a 4-year journey, we are super happy to see this paper out in @NatureComms - @ElezKatarina et al: High-throughput molecular dynamics + active @machinelearning enable efficient identification of an experimentally-validated broad coronavirus inhibitor. https://t.co/Q3ioRNKnJP
nature.com
Nature Communications - Approaches making virtual and experimental screening more resource-efficient are vital for identifying effective inhibitors from a vast pool of potential drugs but remain...
2K
41
257
Really excited to (finally) share the updated JAMUN preprint and codebase! We perform Langevin molecular dynamics in a smoothed space which allows us to take larger integrator steps. This requires learning a score function only at a single noise level, unlike diffusion models.
3
5
45
Our development of machine-learned transferable coarse-grained models in now on Nat Chem! https://t.co/HGngd8Vpop I am so proud of my group for this work! Particularly first authors Nick Charron, Klara Bonneau, @sayeg84, Andrea Guljas.
nature.com
Nature Chemistry - The development of a universal protein coarse-grained model has been a long-standing challenge. A coarse-grained model with chemical transferability has now been developed by...
3
39
122
Come check out SBG happening now! W-115 11-1:30 with @charliebtan
@bose_joey Chen Lin @leonklein26
@mmbronstein
0
15
91
Main Conference: ๐ Title: Scalable Equilibrium Sampling with Sequential Boltzmann Generators ๐ When: Wed 16 Jul 11 a.m. PDT โ 1:30 p.m. PDT ๐บ๏ธ Where: West Exhibition Hall B2-B3 W-115 ๐ arXiv: https://t.co/VKQluYOffP w/ @charliebtan @WillLin1028 @leonklein26 @mmbronstein
arxiv.org
Scalable sampling of molecular states in thermodynamic equilibrium is a long-standing challenge in statistical physics. Boltzmann generators tackle this problem by pairing normalizing flows with...
1
2
7
๐ I'm at #ICML2025 this week, presenting several papers throughout the week with my awesome collaborators! Please do reach out if you'd like to grab a coffee โ๏ธ or catch up again! Papers in ๐งตbelow ๐:
1
7
62
๐Personal update: I'm thrilled to announce that I'm joining Imperial College London @imperialcollege as an Assistant Professor of Computing @ICComputing starting January 2026. My future lab and I will continue to work on building better Generative Models ๐ค, the hardest
98
33
626
New pre-print from PhD student Hang Zou on warm-starting the variational quantum eigensolver using flows: Flow-VQE! Flow-VQE is parameter transfer on steroids: it learns how to solve a family of related problems, dramatically reducing the aggregate compute cost!
1
5
48
I am very happy to share Orbformer, a foundation model for wavefunctions using deep QMC that offers a route to tackle strongly correlated quantum states!
arxiv.org
Reliable description of bond breaking remains a major challenge for quantum chemistry due to the multireferential character of the electronic structure in dissociating species. Multireferential...
7
30
92
The abelian state hidden subgroup problem: Learning stabilizer groups and beyond https://t.co/KwK9pGu85o Identifying the #symmetry properties of quantum states is a central theme in quantum information theory and quantum many-body physics. In this work, we investigate quantum
5
10
59
Excited to have contributed to this amazing work by @LVaitl! https://t.co/Ti8pxrH0mu
Ever felt like Boltzmann Generators trained with Flow Matching were doing fine, just not good enough? We slapped Path Gradients on top, and things got better. No extra samples, no extra compute, no changes to the model. Just gradients you already have access to.
1
1
18
๐ Excited to release a major update to the Boltz-1 model: Boltz-1x! Boltz-1x introduces inference-time steering for much higher physical quality, CUDA kernels for faster, more memory-efficient inference and training, and more! ๐ฅ๐งต
6
109
439
Equivariant functions (e.g. GNNs) can't break symmetries, which can be problematic for generative models and beyond. Come to poster #207 Saturday at 10AM to hear about our solution: SymPE, or symmetry-breaking positional encodings! w/Vasco Portilheiro, Yan Zhang, @sekoumarkaba
2
15
99
Happy to finally release our work on "Composing Unbalanced Flows for Flexible Docking and Relaxation" (FlexDock) that we will be presenting as an oral at #ICLR2025 ! ๐คโ๏ธ๐ธ๐ฌ A thread! ๐งต
2
33
184
Check out cool new work from our group in collaboration with Pfizer and AstraZeneca, lead by Julian Cremer and Ross Irwin on FLOWR, a flow-based ligand generation approach, and highly sanitized benchmark dataset, SPINDR, for the SBDD community!
FLOWR โ Flow Matching for Structure-Aware de novo and Conditional Ligand Generation 1. FLOWR introduces a new generative framework for structure-based ligand design using flow matching instead of diffusion, achieving up to 70x faster inference while improving ligand validity,
2
13
55
Simulating full quantum mechanical ground- and excited state surfaces with deep quantum Monte Carlo by Zeno Schรคtzle, Bernat Szabo and Alice Cuzzocrea. https://t.co/UEy1q3LmLC ๐งตโฌ๏ธ
3
24
96
New preprint! ๐จ We scale equilibrium sampling to hexapeptide (in cartesian coordinates!) with Sequential Boltzmann generators!ย ๐ ๐คฏ Work with @bose_joey, @WillLin1028, @leonklein26, @mmbronstein and @AlexanderTong7 Thread ๐งต 1/11
3
22
76
The BioEmu-1 model and inference code are now public under MIT license!!! Please go ahead, play with it and let us know if there are issues. https://t.co/K7wwHmCt2o
github.com
Inference code for scalable emulation of protein equilibrium ensembles with generative deep learning - microsoft/bioemu
Super excited to preprint our work on developing a Biomolecular Emulator (BioEmu): Scalable emulation of protein equilibrium ensembles with generative deep learning from @MSFTResearch AI for Science. #ML #AI #NeuralNetworks #Biology #AI4Science
https://t.co/yzOy6tAoPv
5
95
356
Check out our new preprint on improving sampling in masked diffusion models! A drop in replacement of the standard random sampling order without any additional training improves performance across the board. ๐
New Paper Alert! ๐ We introduce Path Planning (P2), a sampling approach to optimizing token unmasking order in Masked Diffusion Models (MDMs). SOTA results across language, math, code, and biological sequence (Protein and RNA)๏ฟฝ๏ฟฝall without training. https://t.co/PIsgCS1Hqg ๐งต๐
1
11
82