I can't* fathom why the top picture, and not the bottom picture, is the standard diagram for an autoencoder.
The whole idea of an autoencoder is that you complete a round trip and seek cycle consistency—why lay out the network linearly?
The whole idea of an autoencoder is that you complete a round trip and seek cycle consistency—why lay out the network linearly?
August 29, 2025 at 10:46 PM
I can't* fathom why the top picture, and not the bottom picture, is the standard diagram for an autoencoder.
The whole idea of an autoencoder is that you complete a round trip and seek cycle consistency—why lay out the network linearly?
The whole idea of an autoencoder is that you complete a round trip and seek cycle consistency—why lay out the network linearly?
the best variational autoencoder for rendering fine linework out of a diffusion model was originally produced for hentai
November 9, 2024 at 6:18 AM
the best variational autoencoder for rendering fine linework out of a diffusion model was originally produced for hentai
this is known as an autoencoder
April 12, 2026 at 8:04 PM
this is known as an autoencoder
ok so the memory adapter really wants a better autoencoder so i'm now taking a detour into training that apparently
April 28, 2026 at 3:07 AM
ok so the memory adapter really wants a better autoencoder so i'm now taking a detour into training that apparently
Girls be like "I know a spot" then take you here
June 21, 2025 at 10:00 PM
Girls be like "I know a spot" then take you here
getting tfed into a variational autoencoder
April 10, 2026 at 2:52 AM
getting tfed into a variational autoencoder
you can make an autoencoder out of anything to turn it into numbers and back
April 21, 2026 at 5:10 AM
you can make an autoencoder out of anything to turn it into numbers and back
the model i built is more or less intended to be an instantiation of the vibe autoencoder described in his Glass Bead article.
May 21, 2025 at 4:35 PM
the model i built is more or less intended to be an instantiation of the vibe autoencoder described in his Glass Bead article.
I mean even I made an autoencoder and I got a C in A-level maths
August 8, 2026 at 11:38 AM
I mean even I made an autoencoder and I got a C in A-level maths
December 9, 2024 at 9:04 PM
www.neuronpedia.org/llama3.3-70b...
you can now play with natural language autoencoders right in your browser
you can now play with natural language autoencoders right in your browser
Natural Language Autoencoders – Llama3.3-70B-IT
Natural Language Autoencoder for Llama3.3-70B-IT. Translate a model's internal thoughts into natural language explanations.
www.neuronpedia.org
May 18, 2026 at 12:20 AM
www.neuronpedia.org/llama3.3-70b...
you can now play with natural language autoencoders right in your browser
you can now play with natural language autoencoders right in your browser
“Everyone knows” what an autoencoder is… but there's an important complementary picture missing from most introductory material.
In short: we emphasize how autoencoders are implemented—but not always what they represent (and some of the implications of that representation).🧵
In short: we emphasize how autoencoders are implemented—but not always what they represent (and some of the implications of that representation).🧵
September 6, 2025 at 9:20 PM
“Everyone knows” what an autoencoder is… but there's an important complementary picture missing from most introductory material.
In short: we emphasize how autoencoders are implemented—but not always what they represent (and some of the implications of that representation).🧵
In short: we emphasize how autoencoders are implemented—but not always what they represent (and some of the implications of that representation).🧵
scVAG: Unified single-cell clustering via variational-autoencoder integration with Graph Attention Autoencoder 🧬🖥️ www.sciencedirect.com/science/arti...
scVAG: Unified single-cell clustering via variational-autoencoder integration with Graph Attention Autoencoder
Single-cell RNA sequencing (scRNA-seq) enables high-resolution transcriptional profiling of cell heterogeneity. However, analyzing this noisy, high-di…
www.sciencedirect.com
December 1, 2024 at 5:03 PM
scVAG: Unified single-cell clustering via variational-autoencoder integration with Graph Attention Autoencoder 🧬🖥️ www.sciencedirect.com/science/arti...
this whole model is sandwiched between the encoder and decoder of a variational autoencoder. it would probably work even in pixel space, but the reason it would probably work is less clean than the autoencoder explanation.
bsky.app/profile/theo...
bsky.app/profile/theo...
yeah. these are all samples from the distribution "man, tsdf," where "tsdf" is a purely synthetic label meant to encode everything not captured by a description of what the image entails.
June 9, 2026 at 5:04 AM
this whole model is sandwiched between the encoder and decoder of a variational autoencoder. it would probably work even in pixel space, but the reason it would probably work is less clean than the autoencoder explanation.
bsky.app/profile/theo...
bsky.app/profile/theo...
wait Peli's stuff came *before* your autoencoder? wow
May 21, 2025 at 5:02 PM
wait Peli's stuff came *before* your autoencoder? wow
Replace Variational Autoencoder (VAE) with pretrained representation encoders (e.g., DINO, SigLIP, MAE) paired with trained decoders, which they terms as Representation Autoencoders (RAE).
October 15, 2025 at 3:49 AM
Replace Variational Autoencoder (VAE) with pretrained representation encoders (e.g., DINO, SigLIP, MAE) paired with trained decoders, which they terms as Representation Autoencoders (RAE).
Towards AI-designed genomes using a variational autoencoder https://www.biorxiv.org/content/10.1101/2023.10.22.563484v4 🧬🖥️🧪 https://github.com/natasha-dudek/DeepGenomeVector
November 18, 2024 at 4:30 PM
Towards AI-designed genomes using a variational autoencoder https://www.biorxiv.org/content/10.1101/2023.10.22.563484v4 🧬🖥️🧪 https://github.com/natasha-dudek/DeepGenomeVector
November 16, 2024 at 5:55 PM
you have no idea how much qwen hates being an autoencoder
May 6, 2026 at 9:11 PM
you have no idea how much qwen hates being an autoencoder
i think that the explanation for this is that if you have randomized weights, the sparse autoencoder you train is itself is just a language model. there's no cryptext/codebook distinction.
I Hate Machine Learning I Hate Machine Learning I Hate Machine Learning
arxiv.org/html/2501.17...
arxiv.org/html/2501.17...
Sparse Autoencoders Can Interpret Randomly Initialized Transformers
arxiv.org
March 27, 2025 at 6:27 PM
i think that the explanation for this is that if you have randomized weights, the sparse autoencoder you train is itself is just a language model. there's no cryptext/codebook distinction.
spent a week trying to figure out why the autoencoder was breaking as soon as i try giving it longer passages no matter what i do
...the learning rate was just too low for it -_-
...the learning rate was just too low for it -_-
May 12, 2026 at 6:47 AM
spent a week trying to figure out why the autoencoder was breaking as soon as i try giving it longer passages no matter what i do
...the learning rate was just too low for it -_-
...the learning rate was just too low for it -_-
Microbial intergenic regions encode much of the regulatory and functional logic of the genome, but they remain systematically difficult to discover and annotate.
We trained a sparse autoencoder on gLM2 to identify recurring patterns across hundreds of millions of microbial intergenic regions... 🧵
We trained a sparse autoencoder on gLM2 to identify recurring patterns across hundreds of millions of microbial intergenic regions... 🧵
September 15, 2026 at 5:35 PM
Microbial intergenic regions encode much of the regulatory and functional logic of the genome, but they remain systematically difficult to discover and annotate.
We trained a sparse autoencoder on gLM2 to identify recurring patterns across hundreds of millions of microbial intergenic regions... 🧵
We trained a sparse autoencoder on gLM2 to identify recurring patterns across hundreds of millions of microbial intergenic regions... 🧵
PISCES: Physics-Informed Solar-wind Convolutional autoEncoder for Space-weather Anomaly Detection and Early Warning. Kevin Lee et. al. https://arxiv.org/abs/2609.28022
September 25, 2026 at 8:22 AM
PISCES: Physics-Informed Solar-wind Convolutional autoEncoder for Space-weather Anomaly Detection and Early Warning. Kevin Lee et. al. https://arxiv.org/abs/2609.28022
this is RADE v1, the Radio Autoencoder, via the newly released FreeDV 2.0
June 12, 2025 at 6:49 PM
this is RADE v1, the Radio Autoencoder, via the newly released FreeDV 2.0