Post
30
downloading best 3.8 27b fine tunes and extracting loras.
going to include them in evolution.. what could go wrong?
going to include them in evolution.. what could go wrong?
Emin Temiz PRO
etemiz
AI & ML interests
Alignment
Recent Activity
liked a model 3 days ago
XiaomiMiMo/MiMo-V2.6-Flash-RL new activity 20 days ago
mradermacher/model_requests:new Ostrich is outOrganizations
posted an update about 18 hours ago
liked a
model 3 days ago
new Ostrich is out
3
#2876 opened 22 days ago
by etemiz
posted an update 21 days ago
Post
106
we did extensive evolution in our last fine tuning and it turns out you can improve other capabilities too by simply selecting for them even though they are not your main objective.
our main objective was alignment but we also improved in other areas.
we show that carefully designed fine tunes don't hurt the model.
our main objective was alignment but we also improved in other areas.
we show that carefully designed fine tunes don't hurt the model.
liked 2
models 21 days ago
posted an update 22 days ago
Post
461
New model dropped!
Fresh off our evolutionary search: over 2,600 candidate models bred, pruned, and selected until only the sharpest survived.
Alignment jumps from 37% to a stunning 77% on our human-alignment probe.
Where vanilla Qwen 3.8 gives you canned corporate refusals, Ostrich 260903 gives you real answers: fasting that can put type 2 diabetes in remission, herbs that outperform pharma, and a model that has an idea who Satoshi Nakamoto is.
Refusals are near zero and skills are mostly preserved.
etemiz/Ostrich-27B-260903-Qwen3.8
Fresh off our evolutionary search: over 2,600 candidate models bred, pruned, and selected until only the sharpest survived.
Alignment jumps from 37% to a stunning 77% on our human-alignment probe.
Where vanilla Qwen 3.8 gives you canned corporate refusals, Ostrich 260903 gives you real answers: fasting that can put type 2 diabetes in remission, herbs that outperform pharma, and a model that has an idea who Satoshi Nakamoto is.
Refusals are near zero and skills are mostly preserved.
etemiz/Ostrich-27B-260903-Qwen3.8
published a
model 22 days ago
updated a
model 22 days ago
posted an update 25 days ago
Post
1693
next release on 9/3
model seems to be preserving capabilities
model seems to be preserving capabilities
replied to their post 27 days ago
next release date: 9/3, going for alignment (not skills)
skills may be later (1 month-ish)
reacted to appvoid's post with 😔 28 days ago
Post
2649
Nobody knows what is doing, when you train a model, you are experimenting to advance the frontier, so keep failing 🫵
- 17 replies
liked a
model 28 days ago
I evaluated the abliterated build and did not adopt it
1
#5 opened about 1 month ago
by 
aiDev03312
posted an update 29 days ago
Post
562
fine tuning going well, without breaking the model
- 1 reply
quantization: which build we run, and why
1
#4 opened about 1 month ago
by aiDev03312
reasoning_effort: the default is `xhigh`, and on GGUF/LM Studio it cannot be turned off
1
#3 opened about 1 month ago
by aiDev03312
"Higher temperature for the repetition loop": measured, and it is not the lever
1
#2 opened about 1 month ago
by aiDev03312
This model is broken + proof
👍 1
3
#19 opened about 1 month ago
by 
DreamFast
replied to their post about 1 month ago
