_ - \.
@crumb.bsky.social
lauren (or crumb) // machine // She-E-Ey
hf.co/crumb
hf.co/crumb
Pinned
it's supposed to be, like, a bug
November 11, 2025 at 8:03 PM
it's supposed to be, like, a bug
high pass@k is awesome cause if you actually care about solving problems and getting the best possible solutions it is actually relevant but if you only care about a "product" then obviously it's not worth your time to think about
November 11, 2025 at 8:01 PM
high pass@k is awesome cause if you actually care about solving problems and getting the best possible solutions it is actually relevant but if you only care about a "product" then obviously it's not worth your time to think about
i hope everyone that had a hand in making assistants the norm for what "language models" are goes to hell no matter what
October 2, 2025 at 4:16 PM
i hope everyone that had a hand in making assistants the norm for what "language models" are goes to hell no matter what
have been revisiting this a lot
youtu.be/0BVM0UC28nY
youtu.be/0BVM0UC28nY
September 30, 2025 at 1:43 AM
have been revisiting this a lot
youtu.be/0BVM0UC28nY
youtu.be/0BVM0UC28nY
friggin massive shout out to openinference hosting deepseek v3.1 on openrouter for free
even tho we trained on filtered data generated by deepseek v3 base, our desc2doc model didn't follow prompts as well as we'd hoped. so last night i pounded out a rubric based trainer using deepseek v3.1 (:free) as judge. it is now running. yaaay
September 29, 2025 at 7:05 PM
friggin massive shout out to openinference hosting deepseek v3.1 on openrouter for free
even tho we trained on filtered data generated by deepseek v3 base, our desc2doc model didn't follow prompts as well as we'd hoped. so last night i pounded out a rubric based trainer using deepseek v3.1 (:free) as judge. it is now running. yaaay
September 29, 2025 at 7:05 PM
even tho we trained on filtered data generated by deepseek v3 base, our desc2doc model didn't follow prompts as well as we'd hoped. so last night i pounded out a rubric based trainer using deepseek v3.1 (:free) as judge. it is now running. yaaay
took you long enough Dumb Ass
September 18, 2025 at 3:29 AM
took you long enough Dumb Ass
i think... working towards a set goal like "agi" is not really conducive to finding out what this specific tech stack could be the best at
September 16, 2025 at 6:52 PM
i think... working towards a set goal like "agi" is not really conducive to finding out what this specific tech stack could be the best at
Check out some visualizers like this here:
midwestern-simulation.neocities.org/main/library...
Check out the embedding model we created for them here:
hf.co/midwestern-s...
midwestern-simulation.neocities.org/main/library...
Check out the embedding model we created for them here:
hf.co/midwestern-s...
September 16, 2025 at 5:49 PM
Check out some visualizers like this here:
midwestern-simulation.neocities.org/main/library...
Check out the embedding model we created for them here:
hf.co/midwestern-s...
midwestern-simulation.neocities.org/main/library...
Check out the embedding model we created for them here:
hf.co/midwestern-s...
12 embedding tokens seems to be a sweet spot between reconstruction quality and ability to do math to the embeddings before decoding for our 3b model
September 15, 2025 at 7:21 AM
12 embedding tokens seems to be a sweet spot between reconstruction quality and ability to do math to the embeddings before decoding for our 3b model
we r gonna post like 10 of these in a little bloggy thing to show off the latest essence 3b when it is done training.. more toys... more toys.....
September 11, 2025 at 7:33 PM
we r gonna post like 10 of these in a little bloggy thing to show off the latest essence 3b when it is done training.. more toys... more toys.....
lets go man fuck em up 𝔱𝔬𝔲𝔤𝔥-𝔡𝔯𝔞𝔤𝔬𝔫-₂₅₈
ETA83:50:08
ETA83:50:08
September 3, 2025 at 11:33 PM
lets go man fuck em up 𝔱𝔬𝔲𝔤𝔥-𝔡𝔯𝔞𝔤𝔬𝔫-₂₅₈
ETA83:50:08
ETA83:50:08
subtracting "lamb" embed from mary had a little lamb embed then decoding... it tries to say it but it just cant get it right... that's so silly...
September 2, 2025 at 5:17 AM
subtracting "lamb" embed from mary had a little lamb embed then decoding... it tries to say it but it just cant get it right... that's so silly...
okokokok it's on HF as it is RN, it seems really good but it will keep on improving for a little while,
encourage you to try it out and see if you can figure out any fun things to use it for
hf.co/crumb/essenc...
encourage you to try it out and see if you can figure out any fun things to use it for
hf.co/crumb/essenc...
September 2, 2025 at 4:40 AM
okokokok it's on HF as it is RN, it seems really good but it will keep on improving for a little while,
encourage you to try it out and see if you can figure out any fun things to use it for
hf.co/crumb/essenc...
encourage you to try it out and see if you can figure out any fun things to use it for
hf.co/crumb/essenc...
it apparently generalizes to any number of embedding tokens for any level of detail, from only training on 4,8,16,32,64... even inferencing at 256 doesn't show total degeneration, same w odd nums like 19
what crumb is hoping is the coolest use case is turning any text-in text-out system into a reservoir computer (need to train a VAE on the embeddings first)
this one is for the freaks, have u ever wanted a text2vec2text that 1 doesn't rely on api embeddings and 2 preserves temporal dynamics by design?
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
September 1, 2025 at 5:41 PM
it apparently generalizes to any number of embedding tokens for any level of detail, from only training on 4,8,16,32,64... even inferencing at 256 doesn't show total degeneration, same w odd nums like 19
what crumb is hoping is the coolest use case is turning any text-in text-out system into a reservoir computer (need to train a VAE on the embeddings first)
this one is for the freaks, have u ever wanted a text2vec2text that 1 doesn't rely on api embeddings and 2 preserves temporal dynamics by design?
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
August 28, 2025 at 6:48 PM
what crumb is hoping is the coolest use case is turning any text-in text-out system into a reservoir computer (need to train a VAE on the embeddings first)
trying 2 figure out things to test... like... add embed of structured text (json) to unstructured text, will it structure it? you could jitter an embed a bit to get synthetic data super close to the original? what if you subtracted mean embed of a char's lines from a script, does it remove the char?
this one is for the freaks, have u ever wanted a text2vec2text that 1 doesn't rely on api embeddings and 2 preserves temporal dynamics by design?
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
August 28, 2025 at 6:46 PM
trying 2 figure out things to test... like... add embed of structured text (json) to unstructured text, will it structure it? you could jitter an embed a bit to get synthetic data super close to the original? what if you subtracted mean embed of a char's lines from a script, does it remove the char?
this one is for the freaks, have u ever wanted a text2vec2text that 1 doesn't rely on api embeddings and 2 preserves temporal dynamics by design?
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
August 28, 2025 at 6:33 PM
this one is for the freaks, have u ever wanted a text2vec2text that 1 doesn't rely on api embeddings and 2 preserves temporal dynamics by design?
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
crumb has found crumbself in a position in need of some of these, so crumb is jst building them. 32 token embedding. total 6b model system (WIP results)
Reposted by _ - \.
Oddly good heuristic for what it’s safe to give AI control of
you wouldn't let a demon or fae have access to your bank account either
August 25, 2025 at 3:46 PM
Oddly good heuristic for what it’s safe to give AI control of
crumb found a trove of stuff crumb generated in 2019
August 25, 2025 at 3:36 PM
crumb found a trove of stuff crumb generated in 2019