Does equivariance matter at scale?
Johann Brehmer, Sönke Behrends, Pim De Haan, Taco Cohen
Action editor: Marcus Brubaker
https://openreview.net/forum?id=wilNute8Tn
#models #equivariance #equivariant
New #J2C Certification:
Optimization Dynamics of Equivariant and Augmented Neural Networks
Oskar Nordenfors, Fredrik Ohlsson, Axel Flinth
https://openreview.net/forum?id=PTTa3U29NR
#layers #equivariant #augmented
Capsule Network Projectors Enable Equivariant and Invariant Learning
New research shows capsule network projectors can achieve both equivariant and invariant learning, expanding AI model flexibility. Read more: getnews.me/capsule-network-projecto... #capsulenetwork #equivariant #invariant
Equivariant Flow Matching for Controlled Molecular Generation
Scientists presented an equivariant flow matching method for molecular generation, respecting rotation and translation symmetries. Posted Oct 2025 (arXiv:2506.18340v3). Read more: getnews.me/equivariant-flow-matchin... #moleculargeneration #equivariant
Adaptive Canonicalization Boosts Stability in Equivariant AI Models
Adaptive canonicalization lets models pick the most confident form, improving point‑cloud analysis. Benchmarks report it beats data augmentation and fixed canonicalization. Read more: getnews.me/adaptive-canonicalizatio... #adaptivecanon #equivariant
Clebsch-Gordan Transformer Delivers Fast Global Equivariant Attention
The Clebsch‑Gordan Transformer uses a sparse convolution to keep rotational symmetry and cut complexity to O(N log N), with benchmarks on n‑body showing lower GPU memory. Read more: getnews.me/clebsch-gordan-transform... #clebschgordan #equivariant
New Method Learns Symmetry Groups for Equivariant Neural Networks
A technique learns a quadratic form to discover symmetry groups, enabling equivariant neural nets without preset groups. Tested on polynomial regression and top‑quark tagging. Read more: getnews.me/new-method-learns-symmet... #equivariant #neuralnets
Equivariant AI Surrogate Model for 3D Rayleigh‑Bénard Convection
On Sep 24 2025 researchers released an AI surrogate that combines a G‑steerable autoencoder with a convolutional LSTM to model 3‑D Rayleigh‑Bénard convection. Code on GitHub. Read more: getnews.me/equivariant-ai-surrogate... #equivariant #rayleighbenard
Frame-Based Equivariant Diffusion Advances 3D Molecular Generation
A new frame‑based diffusion model with EdgeDiT sets state‑of‑the‑art results on QM9, with a test negative log‑likelihood of -137.97 and 98.98% atom stability. Read more: getnews.me/frame-based-equivariant-... #qm9 #equivariant
why do #equivariant denoisers struggle with self-symmetrical input, exactly?
The conflicting instructions of #breaking self-symmetry while #maintaining equivariance force an optimally trained denoiser to output the marginal distribution of the node and edge labels in the training dataset.
Diffusion for graphs often uses #equivariant denoisers in order to ensure the model can handle the input in any order. Equivariant denoisers struggle to map a self-symmetrical input into a less self-symmetrical output, which you can see for yourself in this toy notebook: github.com/Aalto-QuML/D...
👉 Are you interested in #diffusion for graphs?
👉 Do you want to know more about the limitations of #equivariant models?
👉 Curious about one of the latest models in #retrosynthesis?
Checkout this 🧵 and come chat with me or
@severi-rissanen.bsky.social anytime at #ICLR2025
Optimization Dynamics of Equivariant and Augmented Neural Networks
Oskar Nordenfors, Fredrik Ohlsson, Axel Flinth
Action editor: Stratis Gavves
https://openreview.net/forum?id=PTTa3U29NR
#layers #equivariant #augmented
Efficient Model-Agnostic Multi-Group Equivariant Networks
Razan Baltaji, Sourya Basu, Lav R. Varshney
Action editor: Jinwoo Shin
https://openreview.net/forum?id=HCMDtc0ZhV
#equivariant #models #groups
The spotlight talk will be on our recent work diving into what gets learned by #equivariant neural networks used for interatomic potentials.
The key finding was that higher order (non-scalar) latent spaces were largely being ignored, which leaves modeling performance on the table.
Decomposition of Equivariant Maps via Invariant Maps: Application to Universal Approximation unde...
Akiyoshi Sannai, Yuuki Takai, Matthieu Cordonnier
Action editor: Alberto Bietti
https://openreview.net/forum?id=ycOLyHh1Ue
#symmetries #invariant #equivariant
