Families — Math in LLMs Research Map

MATH · IN · MODELS

Model families (331)

How many distinct mathematical structures have been found in each family, across every model checkpoint in it. Sorted, tallest first. Click a bar (or a card below) to see the family's models and drill into which structures were found in each one specifically.

27

26

22

17

15

11

10

CLIP (Contrastive Language-Image Pretraining)

9

9

8

8

8

8

7

Show 317 more families

7

7

Custom Research Transformer (small, purpose-built for interpretability studies)

6

6

6

6

Stable Diffusion

6

6

6

Custom Research CNN (purpose-built convolutional net for representation-geometry studies)

5

5

5

5

5

5

Vision Transformer (ViT)

5

4

Custom research feedforward MLP

4

4

4

JEPA (Joint Embedding Predictive Architecture)

4

4

4

OpenAI Embeddings

4

4

4

4

3

Bayesian Wind Tunnel Mamba

3

Bayesian Wind Tunnel Transformer

3

3

CLAP (Contrastive Language-Audio Pretraining)

3

3

3

3

3

3

3

3

3

3

3

3

3

3

3

Weather/Climate Foundation Model

3

3

3

Achilli/Pham et al. Diffusion Memorization-Transition DDPMs

2

2

2

2

2

2

2

2

Chemical Language Model (ChemBERTa / MoLFormer / RoBERTa-Zinc family)

2

2

2

2

2

Custom Recurrent RL Policies

2

DDPM (Denoising Diffusion Probabilistic Models, HuggingFace Diffusers checkpoints)

2

2

2

DIAMOND (diffusion UNet world model)

2

2

2

2

Earth-Observation / Remote-Sensing Foundation Model

2

2

2

2

2

2

2

2

2

2

2

2

2

IRIS (VQ-VAE + Transformer world model)

2

2

Leela Chess Zero

2

2

2

2

2

NLLB (No Language Left Behind)

2

2

2

PGGAN (Progressive GAN)

2

RNNLM (Recurrent Neural Network Language Model)

2

SchNet (continuous-filter convolutional GNN for molecules)

2

Sentiment-classification RNNs (LSTM/GRU/UGRNN/vanilla)

2

SiT (Scalable Interpolant Transformers)

2

2

2

2

2

1

Adaptive Input Transformer LM (Baevski & Auli)

1

ADM (Ablated Diffusion Model, Guided Diffusion)

1

1

1

Autoregressive Predictive Coding (APC)

1

1

Arora et al. SN word embeddings

1

AST (Audio Spectrogram Transformer)

1

Atomistic/Equivariant Graph Neural Network

1

1

1

Banino et al. (2018) grid-cell navigation agent

1

BEiT (BERT Pre-Training of Image Transformers)

1

1

1

Bag-of-Words Superposition Autoencoder (BOWS)

1

BOWS Toy Transformer

1

Brandon, Chadwick & Pellegrino (2025) Boolean-Function MLPs

1

Custom CNN/FC image classifiers (Burfitt, Brodzki & Dłotko 2026)

1

EHR Foundation Model (Llama-3.2-1B architecture, trained on MIMIC-IV)

1

Byte-level multiplicative LSTM (Radford, Józefowicz & Sutskever 2017)

1

CARNN (Continuous Attractor RNN, Sorscher-style path-integration model)

1

1

Sonny (hierarchical weather transformer)

1

ChessGPT (Karvonen)

1

1

Chughtai, Chan & Nanda (2023) group-composition networks

1

Clock/Pizza Modular-Arithmetic Transformer

1

1

1

1

1

Conformal-Normalization Path-Integration RNN

1

Conneau et al. Monolingual MLM

1

Convergent Evolution controlled 300M architecture comparison (Fu, Zhou, Belkin, Sharan & Jia 2026)

1

Counter-Language Transformer

1

Contrastive Predictive Coding (CPC)

1

Cueva & Wei (2018) path-integration RNN

1

Custom Behavior-Cloning Vision Encoder

1

Custom Chess RL Agents

1

Custom Research Contrastive MoE-CNN (sparsely-gated ResNet trained with SimCLR)

1

Custom Molecular Transformer-VAE (purpose-built for SELFIES generation)

1

Custom research sparse mixture-of-experts transformer

1

Custom Research ResNet (constant-width residual variant, purpose-built for SSL studies)

1

Custom Research VAE (purpose-built variational autoencoder for representation-geometry studies)

1

DAC (Descript Audio Codec)

1

1

DCGAN (Deep Convolutional GAN)

1

DDPM++ (Score-Based SDE Diffusion)

1

1

1

1

DRC (Deep Repeating ConvLSTM)

1

Echo State Network / Reservoir Computer

1

1

1

1

1

ESM Cambrian (ESMC)

1

Eternis Forecaster

1

1

1

1

1

Filus et al. (2026) Continual-Learning ResNet-18 (Mammoth framework)

1

1

Gao et al. (2019) tied-embedding NLG models

1

Graph Attention Network (Veličković et al.)

1

Graph Convolutional Network (Kipf & Welling)

1

GELU Toy Transformer

1

1

1

GNN diagnostic-classifier probing suite (GCN/GAT/GIN on Grid-House and ClinTox)

1

Google Massively Multilingual NMT

1

GoogLeNet (Inception v1)

1

Gaussian Process Latent Variable Model (GP-LVM)

1

1

GPT2-base-french

1

Granite Embedding

1

1

Grid-walker toy transformer

1

Grokking Modular-Arithmetic MLP

1

Grokking Modular-Arithmetic Transformer

1

GTE (General Text Embeddings)

1

GTR (Generalizable T5-based dense Retriever)

1

Handke et al. (2025) Custom Class-Conditional DDPMs

1

Haputhanthri, Storan, Jiang et al. (2025) Task-Trained Memory RNNs

1

1

Hierarchical-Equality MLP (custom, Geiger et al. 2023)

1

1

1

1

1

Hyperbolic Graph Neural Networks (HGCN / HyboNet)

1

I3D (Inflated 3D ConvNet)

1

Improved DDPM (iDDPM)

1

1

1

Impala-CNN Atari encoder

1

1

1

Importance-Weighted Autoencoder (IWAE)

1

1

JSB Chorales VAE

1

Jude & Hennig (2020) T-maze LSTM agent

1

Poker Hand History Transformer (Kamel, Rastogi, Ma, Ranganathan & Zhu)

1

Karczewski et al. (2026) custom molecular diffusion model

1

Apple Health Foundation Models (PPG/Accel)

1

Kim et al. (2025) Custom Continual-Learning Networks

1

1

Kiseleva (2026) Boundary-Aware Quantization Toy Classifiers

1

1

Lanza et al. (2026) Continual-Learning ResNet-18 Feature Extractors

1

1

1

1

Passive object-state world models (Liu & Chen 2026)

1

LLaVA-OneVision

1

Brain-to-Speech Decoder (Conv1D + BiGRU + HiFi-GAN, trained on VOCALMIND sEEG)

1

Mannes (2026) Positional-Encoding ViT-S Ablation Suite

1

Custom Research Matrix-Factorization Recommender (purpose-built dot-product recommender for embedding-geometry studies)

1

Maze-Solving Transformer

1

McCracken, Moisescu-Pareja, Letourneau, Precup & Love (2025) modular-addition networks

1

Minegishi, Feng, Furuta et al. (2026) Analogical-Reasoning Toy Transformer

1

1

Controlled Geospatial SSL Encoders (DINO/MAE/MoCo on SSL4EO Sentinel-1/2)

1

1

1

1

1

1

NanoGPT Causal-Attention, No Positional Encoding (Causal-NoPE)

1

NeuralDaredevil

1

NFNet (Normalizer-Free Networks)

1

Nucleotide Transformer

1

1

1

1

1

Pellegrino & Chadwick (2025) Task-Trained Continuous-Time RNNs

1

1

Pi (Physical Intelligence VLA)

1

1

Principal Odor Map GNN

1

Poolside Laguna

1

ProSoRo (Proprioceptive Soft Robot)

1

1

Rangamani (2025) Modular-Addition RNN

1

Reddy (2026) custom toy RNNs (Elman/GRU/LSTM, repeat-copy task)

1

1

1

ruBERT / ruRoBERTa (ai-forever)

1

1

Saanum et al. (2024) Custom Contrastive World Models

1

Saito & Matsubara (2025) Custom MNIST Diffusion Model

1

Saponati et al. (2025) Custom Encoder/Decoder Transformers

1

1

Dolph2Vec (Wav2Vec2-style SSL model for dolphin vocalizations)

1

Shai et al. (2024) Custom Epsilon-Machine Transformers

1

Ordinal Local-Comparison Toy Transformer

1

1

1

1

1

SpeechTokenizer

1

1

SPOT (Self-Training with Patch-Order Permutation)

1

1

1

ST-GCN (Spatio-Temporal Graph Convolutional Network)

1

1

Sudoku Transformer

1

Sussillo & Barak (2013) trained toy RNNs

1

Swaroop (2026) modular-arithmetic ReLU MLP

1

1

TAPE (Tasks Assessing Protein Embeddings)

1

Tian (2024/2025) Quadratic-Activation Toy MLP

1

1

TinyModel (TinyStories-trained)

1

Toy Superposition Autoencoder (Elhage et al.)

1

1

1

Vanilla tanh RNN (task-trained, N=100 neurons)

1

1

Vemula (2026) Custom Decoder-Only Transformers (FineWeb-Edu, from scratch)

1

Ventura et al. (2025) Custom Score-Based Diffusion Models

1

1

1

Voita et al. (2019) objective-matched Transformers (MT/LM/MLM)

1

1

FNN (Foundation Model of Neural Activity)

1

1

Tolman-Eichenbaum Machine (TEM)

1

Bach WTC Autoencoder

1

Wu, Geiger & Millière (2025) Variable-Binding Toy Transformer

1

Euclid Q1 Galaxy Morphology Models (Zoobot ConvNeXt-Nano + self-supervised ViT-S MAE)

1

1

1

1

Mini-ICL toy transformers (custom, Yan, Yang & Zhong 2026)

1

Synthetic GPT-2-style ICL transformers (custom, Yang, Lin, Lee, Papailiopoulos & Nowak 2025)

1

1

0

Convolutional neural network (generic)

0

Cora Graph Convolutional Network (varied-hyperparameter ensemble)

0

0

Grid Walker RNN baseline

0

Grid Walker Transformer

0

ModernCamemBERT

0

Moschella et al. (2022) image (variational) autoencoder

0

0

0

0

Transformer (generic)

0

27 structures · 1 hypothesis · 35 models · 210 observations

Google DeepMind

26 structures · 3 hypotheses · 29 models · 122 observations

22 structures · 75 models · 131 observations

17 structures · 1 hypothesis · 20 models · 61 observations

15 structures · 10 models · 47 observations

Microsoft Research

11 structures · 13 models · 25 observations

10 structures · 8 models · 40 observations

CLIP (Contrastive Language-Image Pretraining)

9 structures · 1 hypothesis · 9 models · 31 observations

9 structures · 1 hypothesis · 8 models · 18 observations

8 structures · 1 hypothesis · 13 models · 54 observations

8 structures · 1 hypothesis · 8 models · 28 observations

8 structures · 8 models · 14 observations

8 structures · 6 models · 10 observations

7 structures · 1 hypothesis · 5 models · 16 observations

Show 317 more families as cards

7 structures · 1 models · 21 observations

Allen Institute for AI

7 structures · 5 models · 7 observations

Custom Research Transformer (small, purpose-built for interpretability studies)

various (academic)

6 structures · 7 models · 6 observations

6 structures · 1 models · 10 observations

Allen Institute for AI

6 structures · 5 models · 13 observations

6 structures · 7 models · 17 observations

Stable Diffusion

6 structures · 10 models · 21 observations

Visual Geometry Group, Oxford

6 structures · 1 models · 7 observations

6 structures · 3 models · 8 observations

Custom Research CNN (purpose-built convolutional net for representation-geometry studies)

various (academic)

5 structures · 7 models · 5 observations

5 structures · 9 models · 10 observations

AllenNLP / Allen Institute for AI

5 structures · 2 models · 6 observations

5 structures · 2 models · 7 observations

Meta AI (Facebook AI Research)

5 structures · 6 models · 6 observations

5 structures · 3 models · 5 observations

Vision Transformer (ViT)

Google Research

5 structures · 6 models · 12 observations

University of Toronto

4 structures · 2 models · 5 observations

Custom research feedforward MLP

4 structures · 9 models · 7 observations

Stanford NLP Group

4 structures · 3 models · 11 observations

Shanghai AI Lab / OpenGVLab

4 structures · 2 models · 6 observations

JEPA (Joint Embedding Predictive Architecture)

Independent / academic

4 structures · 4 models · 5 observations

CMU / Princeton (Gu & Dao)

4 structures · 3 models · 4 observations

4 structures · 3 models · 5 observations

OpenAI Embeddings

4 structures · 2 models · 4 observations

Google DeepMind

4 structures · 2 models · 6 observations

Google Research

4 structures · 3 models · 7 observations

4 structures · 4 models · 5 observations

3 structures · 1 models · 3 observations

Bayesian Wind Tunnel Mamba

Research (Agarwal, Dalal & Misra 2026; Mamba/SSM architecture (Gu & Dao 2023), custom-trained from scratch per synthetic task)

3 structures · 1 models · 1 observations

Bayesian Wind Tunnel Transformer

Research (Agarwal, Dalal & Misra 2026; custom-trained from scratch, one model per synthetic task with an analytically known Bayesian posterior)

3 structures · 2 models · 1 observations

3 structures · 6 models · 5 observations

CLAP (Contrastive Language-Audio Pretraining)

Multiple (Microsoft/LAION-style CLAP architecture)

3 structures · 3 models · 2 observations

3 structures · 3 models · 4 observations

3 structures · 1 models · 2 observations

3 structures · 1 models · 6 observations

3 structures · 5 models · 7 observations

Black Forest Labs

3 structures · 2 models · 8 observations

3 structures · 3 models · 3 observations

various (academic)

3 structures · 6 models · 7 observations

3 structures · 2 models · 5 observations

3 structures · 2 models · 5 observations

Research (Li et al. 2023, custom-trained on synthetic Othello move sequences)

3 structures · 1 models · 4 observations

University of Toronto / Broad Institute (open-source)

3 structures · 1 models · 5 observations

3 structures · 4 models · 9 observations

MCG-NJU (Nanjing University)

3 structures · 2 models · 6 observations

Alibaba (Wan-Video team)

3 structures · 3 models · 3 observations

Weather/Climate Foundation Model

various (Microsoft, Google DeepMind)

3 structures · 2 models · 3 observations

3 structures · 2 models · 5 observations

3 structures · 3 models · 2 observations

Achilli/Pham et al. Diffusion Memorization-Transition DDPMs

Research (Achilli, Ventura, Pham, Raya, Krotov, Lucibello, Ambrogioni et al.; PixelCNN++-backbone DDPMs, arXiv:2410.08727 / arXiv:2505.21777)

2 structures · 5 models · 2 observations

2 structures · 1 models · 1 observations

2 structures · 2 models · 3 observations

Google DeepMind

2 structures · 2 models · 2 observations

2 structures · 2 models · 4 observations

Meta AI (Facebook AI Research)

2 structures · 2 models · 5 observations

2 structures · 4 models · 4 observations

2 structures · 2 models · 2 observations

Chemical Language Model (ChemBERTa / MoLFormer / RoBERTa-Zinc family)

various (academic)

2 structures · 3 models · 2 observations

2 structures · 1 models · 1 observations

Zhipu AI / THUDM

2 structures · 3 models · 3 observations

2 structures · 1 models · 3 observations

2 structures · 2 models · 2 observations

Custom Recurrent RL Policies

2 structures · 3 models · 2 observations

DDPM (Denoising Diffusion Probabilistic Models, HuggingFace Diffusers checkpoints)

Ho, Jain & Abbeel / HuggingFace

2 structures · 10 models · 4 observations

2 structures · 2 models · 3 observations

Huang, Liu, van der Maaten & Weinberger (academic)

2 structures · 2 models · 2 observations

DIAMOND (diffusion UNet world model)

Independent / academic (Alonso et al.)

2 structures · 1 models · 2 observations

2 structures · 2 models · 3 observations

2 structures · 1 models · 3 observations

2 structures · 5 models · 5 observations

Earth-Observation / Remote-Sensing Foundation Model

2 structures · 1 models · 1 observations

2 structures · 2 models · 2 observations

2 structures · 1 models · 3 observations

2 structures · 2 models · 1 observations

Facebook AI Research

2 structures · 3 models · 6 observations

Prakash et al. (2024)

2 structures · 1 models · 1 observations

2 structures · 1 models · 1 observations

Broad Institute

2 structures · 1 models · 2 observations

2 structures · 2 models · 2 observations

2 structures · 1 models · 5 observations

DCC Universidad de Chile

2 structures · 1 models · 2 observations

ByteDance / Academic

2 structures · 2 models · 3 observations

2 structures · 1 models · 2 observations

IRIS (VQ-VAE + Transformer world model)

Independent / academic (Micheli et al.)

2 structures · 1 models · 2 observations

2 structures · 1 models · 2 observations

Leela Chess Zero

Open-source community

2 structures · 1 models · 1 observations

Multi-Agent Perception Lab (m-a-p)

2 structures · 1 models · 2 observations

2 structures · 2 models · 2 observations

Amazon (AutoGluon team)

2 structures · 1 models · 1 observations

2 structures · 2 models · 3 observations

NLLB (No Language Left Behind)

2 structures · 1 models · 2 observations

Allen Institute for AI

2 structures · 1 models · 2 observations

Stanford / TRI / UC Berkeley / Google DeepMind (open collaboration)

2 structures · 1 models · 5 observations

PGGAN (Progressive GAN)

2 structures · 3 models · 3 observations

RNNLM (Recurrent Neural Network Language Model)

Independent / academic (Mikolov et al.)

2 structures · 2 models · 2 observations

SchNet (continuous-filter convolutional GNN for molecules)

research/foundational-architecture

2 structures · 1 models · 1 observations

Sentiment-classification RNNs (LSTM/GRU/UGRNN/vanilla)

Maheswaranathan, Williams, Golub, Ganguli & Sussillo (academic)

2 structures · 4 models · 2 observations

SiT (Scalable Interpolant Transformers)

2 structures · 2 models · 2 observations

2 structures · 2 models · 2 observations

2 structures · 4 models · 4 observations

Soda-Inria / academic open-source

2 structures · 1 models · 2 observations

2 structures · 1 models · 3 observations

ACE Studio / StepFun

1 structure · 1 models · 1 observations

Adaptive Input Transformer LM (Baevski & Auli)

Facebook AI Research

1 structure · 1 models · 1 observations

ADM (Ablated Diffusion Model, Guided Diffusion)

1 structure · 1 models · 1 observations

Academic / multi-institution collaboration (Parker, Lanusse, Shen et al., 26 authors incl. Cho, Cranmer, Ho)

1 structure · 1 models · 1 observations

Rostlab / TU Munich

1 structure · 1 models · 1 observations

Autoregressive Predictive Coding (APC)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Arora et al. SN word embeddings

Arora, Li, Liang, Ma & Risteski (2016/2018)

1 structure · 1 models · 1 observations

AST (Audio Spectrogram Transformer)

1 structure · 1 models · 1 observations

Atomistic/Equivariant Graph Neural Network

various (academic)

1 structure · 3 models · 1 observations

various (academic)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Banino et al. (2018) grid-cell navigation agent

DeepMind (Banino et al. 2018)

1 structure · 1 models · 1 observations

BEiT (BERT Pre-Training of Image Transformers)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

SpeakLeash / ACK Cyfronet AGH

1 structure · 1 models · 1 observations

Bag-of-Words Superposition Autoencoder (BOWS)

Research (Prieto, Stevinson, Barsbey, Birdal & Mediano 2026; tied-weight autoencoder, custom-trained from scratch on binary bag-of-words vectors of WikiText-103 / OpenWebText text chunks)

1 structure · 2 models · 1 observations

BOWS Toy Transformer

Research (Prieto, Stevinson, Barsbey, Birdal & Mediano 2026; custom 2-layer causal transformer, d_model=768, 8 heads, GELU FFN width 4*d_model, trained from scratch on WikiText-103 with a multi-hot 'seen-so-far' token-recovery objective)

1 structure · 1 models · 1 observations

Brandon, Chadwick & Pellegrino (2025) Boolean-Function MLPs

Research (Univ. Edinburgh / ENS Paris / UCL Gatsby Unit; custom-trained from scratch, arXiv:2512.00196)

1 structure · 1 models · 1 observations

Custom CNN/FC image classifiers (Burfitt, Brodzki & Dłotko 2026)

1 structure · 2 models · 1 observations

EHR Foundation Model (Llama-3.2-1B architecture, trained on MIMIC-IV)

University of Chicago / Rush University (Burkhart, Ramadan, Liao, Chhikara, Rojas, Parker & Beaulieu-Jones)

1 structure · 1 models · 1 observations

Byte-level multiplicative LSTM (Radford, Józefowicz & Sutskever 2017)

OpenAI (Radford, Józefowicz & Sutskever 2017, trained on Amazon reviews)

1 structure · 1 models · 1 observations

CARNN (Continuous Attractor RNN, Sorscher-style path-integration model)

Schøyen et al. (academic, University of Oslo/Simula)

1 structure · 1 models · 1 observations

Zhipu AI / Tsinghua KEG

1 structure · 1 models · 1 observations

Sonny (hierarchical weather transformer)

Sejong University (Minjong Cheon)

1 structure · 1 models · 1 observations

ChessGPT (Karvonen)

Independent (Adam Karvonen)

1 structure · 2 models · 1 observations

1 structure · 1 models · 2 observations

Chughtai, Chan & Nanda (2023) group-composition networks

Chughtai, Chan & Nanda (2023), ICML

1 structure · 1 models · 1 observations

Clock/Pizza Modular-Arithmetic Transformer

Research (Zhong, Liu, Tegmark & Andreas 2023; custom-trained from scratch, large hyperparameter/initialization sweep on modular addition)

1 structure · 2 models · 1 observations

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Salesforce Research

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Conformal-Normalization Path-Integration RNN

Xu, Gao, Zhang, Wei & Wu (academic)

1 structure · 2 models · 1 observations

Conneau et al. Monolingual MLM

Facebook AI / Johns Hopkins (Conneau, Wu, Li, Zettlemoyer & Stoyanov 2020; custom 8-layer bidirectional Transformer MLM, d_model=768, 12 heads, FFN width 3072, GELU, trained from scratch per-language on Wikipedia using the Lample & Conneau (2019) MLM implementation, architecture held fixed across languages for controlled comparison)

1 structure · 5 models · 1 observations

Convergent Evolution controlled 300M architecture comparison (Fu, Zhou, Belkin, Sharan & Jia 2026)

USC / UC San Diego

1 structure · 3 models · 1 observations

Counter-Language Transformer

Research (Tiwari, Gupta & Hahn 2025; architecture and training setup following Bhattamishra, Ahuja & Goyal 2020, custom-trained from scratch per formal language)

1 structure · 1 models · 1 observations

Contrastive Predictive Coding (CPC)

1 structure · 2 models · 1 observations

Cueva & Wei (2018) path-integration RNN

Research (Cueva & Wei 2018, custom-trained vanilla RNN)

1 structure · 1 models · 1 observations

Custom Behavior-Cloning Vision Encoder

academic (various)

1 structure · 1 models · 1 observations

Custom Chess RL Agents

1 structure · 2 models · 1 observations

Custom Research Contrastive MoE-CNN (sparsely-gated ResNet trained with SimCLR)

various (academic)

1 structure · 1 models · 1 observations

Custom Molecular Transformer-VAE (purpose-built for SELFIES generation)

various (academic)

1 structure · 1 models · 1 observations

Custom research sparse mixture-of-experts transformer

1 structure · 2 models · 2 observations

Custom Research ResNet (constant-width residual variant, purpose-built for SSL studies)

various (academic)

1 structure · 1 models · 1 observations

Custom Research VAE (purpose-built variational autoencoder for representation-geometry studies)

various (academic)

1 structure · 5 models · 4 observations

DAC (Descript Audio Codec)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

DCGAN (Deep Convolutional GAN)

Radford, Metz & Chintala (2016), ICLR

1 structure · 1 models · 1 observations

DDPM++ (Score-Based SDE Diffusion)

Song et al. (Stanford)

1 structure · 3 models · 1 observations

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

ASLP@NPU / Xiaomi

1 structure · 1 models · 1 observations

DRC (Deep Repeating ConvLSTM)

Academic (Guez et al. / Garriga-Alonso Lab)

1 structure · 1 models · 2 observations

Echo State Network / Reservoir Computer

various (academic, reservoir-computing tradition)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

NVIDIA (Karras et al.)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

EvolutionaryScale

1 structure · 2 models · 2 observations

ESM Cambrian (ESMC)

EvolutionaryScale

1 structure · 2 models · 2 observations

Eternis Forecaster

1 structure · 1 models · 1 observations

EuroBERT Team (multi-institution EU consortium)

1 structure · 2 models · 1 observations

Unbabel / EuroLLM project

1 structure · 2 models · 2 observations

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Filus et al. (2026) Continual-Learning ResNet-18 (Mammoth framework)

Research (Filus, Faber, Corizzo & Kanan; trained from scratch via the Mammoth CL framework, arXiv:2605.16374)

1 structure · 2 models · 1 observations

1 structure · 1 models · 2 observations

Gao et al. (2019) tied-embedding NLG models

Gao, He, Tan, Qin, Wang & Liu (2019), ICLR

1 structure · 1 models · 1 observations

Graph Attention Network (Veličković et al.)

research/foundational-architecture

1 structure · 1 models · 1 observations

Graph Convolutional Network (Kipf & Welling)

research/foundational-architecture

1 structure · 1 models · 1 observations

GELU Toy Transformer

Neel Nanda (independent/TransformerLens)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

GNN diagnostic-classifier probing suite (GCN/GAT/GIN on Grid-House and ClinTox)

1 structure · 1 models · 1 observations

Google Massively Multilingual NMT

1 structure · 1 models · 1 observations

GoogLeNet (Inception v1)

1 structure · 1 models · 1 observations

Gaussian Process Latent Variable Model (GP-LVM)

various (academic)

1 structure · 2 models · 1 observations

1 structure · 1 models · 1 observations

GPT2-base-french

community (HuggingFace)

1 structure · 1 models · 2 observations

Granite Embedding

1 structure · 1 models · 2 observations

1 structure · 1 models · 1 observations

Grid-walker toy transformer

Brenner, Knosche & Scherf (2026); custom-trained from scratch on 2D-lattice random walks

1 structure · 1 models · 1 observations

Grokking Modular-Arithmetic MLP

Research (custom-trained from scratch on modular addition, MLP architecture as a non-attention grokking comparison point)

1 structure · 1 models · 1 observations

Grokking Modular-Arithmetic Transformer

Research (Nanda, Chan, Lieberum, Smith & Steinhardt 2023; custom-trained from scratch on modular addition and related algorithmic tasks)

1 structure · 2 models · 3 observations

GTE (General Text Embeddings)

1 structure · 2 models · 2 observations

GTR (Generalizable T5-based dense Retriever)

1 structure · 1 models · 3 observations

Handke et al. (2025) Custom Class-Conditional DDPMs

Research (Handke, Koulischer, Raya & Ambrogioni; custom-trained from scratch, arXiv:2506.10433)

1 structure · 1 models · 1 observations

Haputhanthri, Storan, Jiang et al. (2025) Task-Trained Memory RNNs

Research (Stanford CNC / NTT Physics & Informatics Labs; custom-trained from scratch, arXiv:2502.07256)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Hierarchical-Equality MLP (custom, Geiger et al. 2023)

Stanford (custom-trained)

1 structure · 1 models · 1 observations

ELLIS Institute / MPI-IS / University of Maryland (Geiping et al.)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Pal et al. (academic)

1 structure · 1 models · 1 observations

Stanford University

1 structure · 1 models · 1 observations

Hyperbolic Graph Neural Networks (HGCN / HyboNet)

Various (Chami et al. 2019 HGCN; Chen et al. 2022 HyboNet)

1 structure · 2 models · 1 observations

I3D (Inflated 3D ConvNet)

1 structure · 1 models · 1 observations

Improved DDPM (iDDPM)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

1 structure · 3 models · 1 observations

Impala-CNN Atari encoder

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Shanghai AI Laboratory

1 structure · 2 models · 1 observations

Importance-Weighted Autoencoder (IWAE)

various (academic)

1 structure · 3 models · 1 observations

MyShell.ai / Princeton

1 structure · 1 models · 1 observations

JSB Chorales VAE

University of Porto / INESC-TEC

1 structure · 1 models · 1 observations

Jude & Hennig (2020) T-maze LSTM agent

Research (Jude & Hennig 2020, University of Edinburgh)

1 structure · 1 models · 1 observations

Poker Hand History Transformer (Kamel, Rastogi, Ma, Ranganathan & Zhu)

Algoverse AI Research / Waterloo / Berkeley

1 structure · 1 models · 1 observations

Karczewski et al. (2026) custom molecular diffusion model

custom (paper-trained)

1 structure · 1 models · 1 observations

Apple Health Foundation Models (PPG/Accel)

Apple (Katuwal, Koparkar, Abbaspourazad, Mishra & Kirthivasan)

1 structure · 1 models · 1 observations

Kim et al. (2025) Custom Continual-Learning Networks

Research (Kim, Kim & Sohn; custom-trained from scratch, arXiv:2505.20970, ICML 2025)

1 structure · 2 models · 1 observations

1 structure · 1 models · 1 observations

Kiseleva (2026) Boundary-Aware Quantization Toy Classifiers

Research (Kiseleva, Innopolis University, 2026; custom-trained small classifiers under post-training and quantization-aware quantization, arXiv:2607.01478)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Lanza et al. (2026) Continual-Learning ResNet-18 Feature Extractors

Research (Lanza, Pereira, Miozzo, Angelats & Dini; trained from scratch, arXiv:2602.06586)

1 structure · 2 models · 1 observations

classic CNN architecture (LeCun et al.)

1 structure · 1 models · 1 observations

Hong Kong Polytechnic University / UC Berkeley (Lin, He, Mak, Lian & Lee)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Passive object-state world models (Liu & Chen 2026)

Independent / academic (Fudan University)

1 structure · 2 models · 1 observations

LLaVA-OneVision

ByteDance / LMMs-Lab

1 structure · 2 models · 2 observations

Brain-to-Speech Decoder (Conv1D + BiGRU + HiFi-GAN, trained on VOCALMIND sEEG)

University of Maryland College Park (Maghsoudi & Mishra)

1 structure · 1 models · 1 observations

Mannes (2026) Positional-Encoding ViT-S Ablation Suite

Research (Mannes, 2026; custom-trained from scratch on ImageNet-100 across five positional-encoding variants, arXiv:2606.00124)

1 structure · 1 models · 1 observations

Custom Research Matrix-Factorization Recommender (purpose-built dot-product recommender for embedding-geometry studies)

various (academic)

1 structure · 1 models · 1 observations

Maze-Solving Transformer

Research (Ivanitskiy et al. 2023, custom-trained on synthetic maze-solving token sequences)

1 structure · 4 models · 1 observations

McCracken, Moisescu-Pareja, Letourneau, Precup & Love (2025) modular-addition networks

McCracken, Moisescu-Pareja, Letourneau, Precup & Love (2025)

1 structure · 1 models · 1 observations

Minegishi, Feng, Furuta et al. (2026) Analogical-Reasoning Toy Transformer

Research (Minegishi et al., ICML 2026; custom-trained from scratch on synthetic entity-relation analogies, arXiv:2602.01992)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Controlled Geospatial SSL Encoders (DINO/MAE/MoCo on SSL4EO Sentinel-1/2)

Johns Hopkins Applied Physics Lab / Johns Hopkins University

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Allen Institute for AI

1 structure · 1 models · 1 observations

Carnegie Mellon University / Autodesk

1 structure · 1 models · 2 observations

1 structure · 1 models · 1 observations

Google Research

1 structure · 5 models · 1 observations

NanoGPT Causal-Attention, No Positional Encoding (Causal-NoPE)

Research (Zuo, Guerzhoy & Guerzhoy 2025; custom-trained NanoGPT variants with causal attention and no positional encodings, on synthetic position-sensitive tasks)

1 structure · 1 models · 1 observations

NeuralDaredevil

1 structure · 1 models · 1 observations

NFNet (Normalizer-Free Networks)

1 structure · 1 models · 1 observations

Nucleotide Transformer

InstaDeep / NVIDIA / TUM

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

LAION / University of Washington / Stanford / others (open-source)

1 structure · 3 models · 2 observations

1 structure · 1 models · 1 observations

1 structure · 2 models · 1 observations

Pellegrino & Chadwick (2025) Task-Trained Continuous-Time RNNs

Research (Pellegrino & Chadwick, NeurIPS 2025; custom-trained from scratch, arXiv:2512.04310)

1 structure · 2 models · 1 observations

independent research

1 structure · 1 models · 1 observations

Pi (Physical Intelligence VLA)

Physical Intelligence

1 structure · 2 models · 3 observations

PLLuM Consortium

1 structure · 1 models · 1 observations

Principal Odor Map GNN

Google Research / Osmo

1 structure · 1 models · 1 observations

Poolside Laguna

1 structure · 2 models · 1 observations

ProSoRo (Proprioceptive Soft Robot)

Southern University of Science and Technology / Shanghai Jiao Tong University

1 structure · 1 models · 1 observations

RostLab (Technical University of Munich)

1 structure · 4 models · 1 observations

Rangamani (2025) Modular-Addition RNN

Research (Rangamani 2025; custom-trained from scratch on modular addition, arXiv:2503.22059)

1 structure · 1 models · 1 observations

Reddy (2026) custom toy RNNs (Elman/GRU/LSTM, repeat-copy task)

custom (paper-trained)

1 structure · 4 models · 1 observations

1 structure · 1 models · 1 observations

academic (RoboRefer authors)

1 structure · 1 models · 1 observations

ruBERT / ruRoBERTa (ai-forever)

SberDevices / ai-forever

1 structure · 3 models · 1 observations

RWKV community / Recursal

1 structure · 1 models · 1 observations

Saanum et al. (2024) Custom Contrastive World Models

Research (Saanum, Schulze Buschoff, Dayan & Schulz; custom-trained from scratch, arXiv:2410.04940)

1 structure · 2 models · 1 observations

Saito & Matsubara (2025) Custom MNIST Diffusion Model

Research (Saito & Matsubara; custom-trained from scratch on MNIST, arXiv:2504.20288)

1 structure · 1 models · 1 observations

Saponati et al. (2025) Custom Encoder/Decoder Transformers

Research (Saponati, Sager, Vilimelis Aceituno, Stadelmann & Grewe; trained from scratch, arXiv:2502.10927)

1 structure · 2 models · 1 observations

1 structure · 1 models · 1 observations

Dolph2Vec (Wav2Vec2-style SSL model for dolphin vocalizations)

ENS Paris / Institut du Cerveau / Champalimaud Foundation

1 structure · 1 models · 1 observations

Shai et al. (2024) Custom Epsilon-Machine Transformers

Research (Shai, Marzen, Teixeira, Gietelink Oldenziel & Riechers; custom-trained from scratch, arXiv:2405.15943, NeurIPS 2024 spotlight)

1 structure · 1 models · 1 observations

Ordinal Local-Comparison Toy Transformer

Research (Singh 2026; custom-trained from scratch on adjacent-pair comparisons over total orders of N=10/15/20 synthetic entities)

1 structure · 1 models · 1 observations

research community

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

1 structure · 1 models · 2 observations

SpeechTokenizer

Fudan University

1 structure · 1 models · 1 observations

Liu, Wen, Yu, Li, Raj & Song (academic)

1 structure · 1 models · 1 observations

SPOT (Self-Training with Patch-Order Permutation)

research community

1 structure · 1 models · 1 observations

1 structure · 1 models · 2 observations

1 structure · 1 models · 2 observations

ST-GCN (Spatio-Temporal Graph Convolutional Network)

open-source (research)

1 structure · 1 models · 1 observations

various (academic)

1 structure · 1 models · 1 observations

Sudoku Transformer

Research (Kniazev & Fijalkow 2026; architecture following Giannoulis, Pantis & Tzamos 2026, custom-trained from scratch on Norvig-style Sudoku solving traces)

1 structure · 1 models · 1 observations

Sussillo & Barak (2013) trained toy RNNs

Research (Sussillo & Barak 2013, custom-trained continuous-time RNNs)

1 structure · 1 models · 1 observations

Swaroop (2026) modular-arithmetic ReLU MLP

Research (Swaroop 2026; custom-trained from scratch on modular arithmetic)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

TAPE (Tasks Assessing Protein Embeddings)

1 structure · 2 models · 1 observations

Tian (2024/2025) Quadratic-Activation Toy MLP

Research (Tian, NeurIPS 2025; custom-trained from scratch on synthetic modular-addition/Abelian-group tasks, arXiv:2410.01779)

1 structure · 1 models · 1 observations

TinyLlama project

1 structure · 1 models · 1 observations

TinyModel (TinyStories-trained)

Noa Nabeshima (independent)

1 structure · 1 models · 1 observations

Toy Superposition Autoencoder (Elhage et al.)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

Allen Institute for AI

1 structure · 1 models · 1 observations

Vanilla tanh RNN (task-trained, N=100 neurons)

various (academic, computational-neuroscience tradition)

1 structure · 2 models · 1 observations

Renmin University of China / Baidu

1 structure · 2 models · 1 observations

Vemula (2026) Custom Decoder-Only Transformers (FineWeb-Edu, from scratch)

Research (Vemula; custom-trained from scratch, arXiv:2606.20743)

1 structure · 2 models · 1 observations

Ventura et al. (2025) Custom Score-Based Diffusion Models

Research (Ventura, Achilli, Silvestri, Lucibello & Ambrogioni; custom-trained from scratch, arXiv:2410.05898)

1 structure · 3 models · 1 observations

Google DeepMind

1 structure · 1 models · 1 observations

research community

1 structure · 1 models · 1 observations

Voita et al. (2019) objective-matched Transformers (MT/LM/MLM)

Voita, Sennrich & Titov (2019), EMNLP-IJCNLP

1 structure · 3 models · 1 observations

1 structure · 1 models · 1 observations

FNN (Foundation Model of Neural Activity)

Wang et al. (2025, Nature); MICrONS consortium data

1 structure · 1 models · 1 observations

Zhejiang University / Alibaba

1 structure · 1 models · 1 observations

Tolman-Eichenbaum Machine (TEM)

Whittington, Muller, Mark, Chen, Barry, Burgess & Behrens (2020), Cell

1 structure · 1 models · 1 observations

Bach WTC Autoencoder

American University of Beirut

1 structure · 1 models · 1 observations

Wu, Geiger & Millière (2025) Variable-Binding Toy Transformer

Research (Wu, Geiger & Millière, ICML 2025; custom-trained from scratch on synthetic variable-assignment programs, arXiv:2505.20896)

1 structure · 1 models · 1 observations

Euclid Q1 Galaxy Morphology Models (Zoobot ConvNeXt-Nano + self-supervised ViT-S MAE)

STScI / Johns Hopkins / University of Toronto (Wu & Walmsley)

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

1 structure · 1 models · 1 observations

NXAI / JKU Linz

1 structure · 1 models · 1 observations

Mini-ICL toy transformers (custom, Yan, Yang & Zhong 2026)

University of Wisconsin-Madison / University of Chicago (custom-trained)

1 structure · 1 models · 1 observations

Synthetic GPT-2-style ICL transformers (custom, Yang, Lin, Lee, Papailiopoulos & Nowak 2025)

University of Wisconsin-Madison (custom-trained)

1 structure · 1 models · 1 observations

1 structure · 1 models · 2 observations

0 structures · 5 models · 1 observations

Convolutional neural network (generic)

0 structures · 1 models · 2 observations

Cora Graph Convolutional Network (varied-hyperparameter ensemble)

0 structures · 1 models · 1 observations

Microsoft Research

0 structures · 1 models · 1 observations

Grid Walker RNN baseline

Research (Brenner, Knösche & Scherf 2026; LSTM/GRU baselines custom-trained from scratch on the same synthetic 2D grid-walk sequences)

0 structures · 2 models · 0 observations

Grid Walker Transformer

Research (Brenner, Knösche & Scherf 2026; TransformerLens HookedTransformer, custom-trained from scratch on synthetic 2D grid-walk sequences)

0 structures · 2 models · 0 observations

ModernCamemBERT

Inria (ALMAnaCH)

0 structures · 1 models · 0 observations

Moschella et al. (2022) image (variational) autoencoder

Sapienza University of Rome

0 structures · 1 models · 1 observations

OpenLM Research

0 structures · 3 models · 2 observations

Salesforce Research

0 structures · 1 models · 1 observations

0 structures · 1 models · 1 observations

Transformer (generic)

0 structures · 1 models · 1 observations