Beta
Explore
Marketplace
Neural Labs
Chat
Wallet
Docs
qwen3.6-27b-mixture-500k-da20-numina – Dataset by LASR-Callum | AlphaNeural AI
You can deploy this model and start earning money today!
LASR-Callum
/
qwen3.6-27b-mixture-500k-da20-numina
like
0
text-generation
en
odc-by
n<1K
us
sft
qwen3.6
synthdoc-v2
difficult-advice
numinamath
tulu3
no-robots
empty-think
assistant-only-loss
training-data
mixture
qwen3.6-27b-mixture-500k-da20-numina
unknown
final
backfilled-2026-08-25
Views
No views yet
Model card
Files and Versions
Community
API
Qwen3.6-27B SFT mixture — 500k, 20% difficult-advice + maths-weighted remainder
499,344 tokens across 857 conversations. md5 416c72ef854a66d0036566c6e9ccec24.
Source Examples Tokens Share Think block
difficult-advice (synthdoc_v2) 56 95,813 19.19% real reasoning trace
NuminaMath-CoT 492 269,451 53.96% none
No Robots 215 67,345 13.49% empty marker
TULU3 94 66,735 13.36% empty marker
Total 857 499,344
Within the non-difficult-advice 80.8%: NuminaMath… See the full description on the dataset page:
https://huggingface.co/datasets/LASR-Callum/qwen3.6-27b-mixture-500k-da20-numina
.
