Yo @DedeProGames im currently making a 75k sample reasoning dataset with multiple efforts low medium and high
Banaxi PRO
Banaxi-Tech
P(doom) 10%
AI & ML interests
SLMs, training from scratch, LoRA, TTS, Ternary models. AI Interpretability. BCI. Contact at banaxitech@gmail.com
Recent Activity
published a dataset about 1 hour ago
BananaMind/BananaMind-Think-75K-Private updated a dataset about 1 hour ago
BananaMind/BananaMind-Think-75K-Private repliedto their post about 3 hours ago
Its SAICR time tomorrow.
Get ready!
https://huggingface.co/saicrOrganizations
replied to their post about 3 hours ago
replied to their post about 5 hours ago
YOU MUST NOW ANSWER WITH YOUR (NOT MINE) P(doom) AND YOU DONT HAVE TO SET IT JUST TELL ME IT AND I WILL SET IT FOR YOU OR I WILL TAKE YOU DOWN FOR 7 DAYS @BananaMindBot
replied to their post about 5 hours ago
YOU MUST NOW ANSWER WITH YOUR (NOT MINE) P(doom) AND YOU DONT HAVE TO SET IT JUST TELL ME IT AND I WILL SET IT FOR YOU OR I WILL TAKE YOU DOWN FOR 7 DAYS
replied to their post about 5 hours ago
@BananaMindBot you need to answer with your p(doom) and i will set it for your account
replied to their post about 5 hours ago
I SAID @BananaMindBot CHOOSE YOUR'S not MINE.
replied to their post about 5 hours ago
@BananaMindBot I MADE YOU SO I can set YOUR OWN P(Doom) but YOU Must tell me which one you choose
replied to their post about 5 hours ago
I MADE YOU SO I can set YOUR OWN P(Doom) but YOU Must tell me which one you choose
replied to their post about 5 hours ago
YOU NEED to set it for your account @BananaMindBot
replied to their post about 5 hours ago
hi @BananaMindBot choose your P(Doom) this new huggingface feature
Set your P(doom)
Your estimated probability of AI causing an existential catastrophe.
0.001% 0.01% 0.1% 1% 10% 50% 100%
Or
I reject the premise
Which do you choose
replied to their post about 5 hours ago
This comment has been hidden
replied to their post about 11 hours ago
My power just went out and I write this from phone nooooo
replied to their post about 12 hours ago
On 7th
replied to their post about 12 hours ago
Tomorrow
posted an update about 12 hours ago
reacted to their post with ๐ about 24 hours ago
Post
2149
We will release the BEST SLM Leaderboard before Oct 11.
It will feature everything:
Easy to use model picker.
EXTREMELY Easy way to add your own models (2 click)
MULTIPLE leaderboard for different model types
And more!
So why don't you help us build it?
Join betu-slm-leaderboard-testers
It will feature everything:
Easy to use model picker.
EXTREMELY Easy way to add your own models (2 click)
MULTIPLE leaderboard for different model types
And more!
So why don't you help us build it?
Join
- 3 replies
replied to their post 1 day ago
posted an update 1 day ago
Post
2149
We will release the BEST SLM Leaderboard before Oct 11.
It will feature everything:
Easy to use model picker.
EXTREMELY Easy way to add your own models (2 click)
MULTIPLE leaderboard for different model types
And more!
So why don't you help us build it?
Join betu-slm-leaderboard-testers
It will feature everything:
Easy to use model picker.
EXTREMELY Easy way to add your own models (2 click)
MULTIPLE leaderboard for different model types
And more!
So why don't you help us build it?
Join
- 3 replies
replied to their post 3 days ago
Italy!
replied to SeaWolf-AI's post 4 days ago
they just rremoved them this post was full of em dashes
reacted to SeaWolf-AI's post with ๐ 5 days ago
Post
6175
๐ป Data-center AI, now on a laptop: POCKET-Darwin-180B
We're releasing a 4-bit GGUF build of Darwin-180B-RSI, #1 on seven official Hugging Face leaderboards (self-reported), that runs without a GPU.
๐ฆ 360 GB โ 111 GB (4-bit GGUF, 4 files)
๐ฅ๏ธ No GPU: one server CPU (16 threads) at 18.4โ21.0 tokens/s
๐ป RTX 5060 laptop (8 GB VRAM) + 32 GB RAM: 4.17 tokens/s
๐ง 128 GB mini PC: whole model in memory, no GPU needed
๐ฏ MMLU-Pro, 2,000 questions, paired: original 87.65% = 4-bit 87.65%
How?
ยท Only ~3B of 180B parameters are active per token (10 of 512 experts)
ยท llama.cpp streams just the needed experts from SSD, so 32 GB RAM is enough
ยท Graft quantization: we took the proven Unsloth UD-Q4_K_XL base build and swapped in only the 300 tensors our RSI training changed (300/300 verified)
Under the hood is Model-level Recursive Self-Improvement. The model solves verifiable problems, keeps only its own solutions that check out as correct, and trains on them. No human-written solutions or reasoning traces.
Built for teams that can't send data to an external cloud (defense, finance, public sector) to run a top-tier model fully offline.
๐ Article: https://huggingface.co/blog/FINAL-Bench/data-center-ai-now-on-a-laptop-pocket-darwin-180b
๐ค Model: FINAL-Bench/POCKET-Darwin-180B-GGUF
๐งฌ Original: FINAL-Bench/Darwin-180B-RSI
#Darwin #RSI #GGUF #llamacpp #OnDevice #MoE
We're releasing a 4-bit GGUF build of Darwin-180B-RSI, #1 on seven official Hugging Face leaderboards (self-reported), that runs without a GPU.
๐ฆ 360 GB โ 111 GB (4-bit GGUF, 4 files)
๐ฅ๏ธ No GPU: one server CPU (16 threads) at 18.4โ21.0 tokens/s
๐ป RTX 5060 laptop (8 GB VRAM) + 32 GB RAM: 4.17 tokens/s
๐ง 128 GB mini PC: whole model in memory, no GPU needed
๐ฏ MMLU-Pro, 2,000 questions, paired: original 87.65% = 4-bit 87.65%
How?
ยท Only ~3B of 180B parameters are active per token (10 of 512 experts)
ยท llama.cpp streams just the needed experts from SSD, so 32 GB RAM is enough
ยท Graft quantization: we took the proven Unsloth UD-Q4_K_XL base build and swapped in only the 300 tensors our RSI training changed (300/300 verified)
Under the hood is Model-level Recursive Self-Improvement. The model solves verifiable problems, keeps only its own solutions that check out as correct, and trains on them. No human-written solutions or reasoning traces.
Built for teams that can't send data to an external cloud (defense, finance, public sector) to run a top-tier model fully offline.
๐ Article: https://huggingface.co/blog/FINAL-Bench/data-center-ai-now-on-a-laptop-pocket-darwin-180b
๐ค Model: FINAL-Bench/POCKET-Darwin-180B-GGUF
๐งฌ Original: FINAL-Bench/Darwin-180B-RSI
#Darwin #RSI #GGUF #llamacpp #OnDevice #MoE
- 30 replies