Beta
Explore
Marketplace
Neural Labs
Chat
Wallet
Docs
kanitakorn-r1d-qwen14b-v12-education-20260523 – AI Model by Jnx03 | AlphaNeural AI
You can deploy this model and start earning money today!
Jnx03
/
kanitakorn-r1d-qwen14b-v12-education-20260523
like
0
safetensors
thai
education
p1-m6
thaiexam
_edu
kanitakorn
th
en
deepseek-ai/DeepSeek-R1-Distill-Qwen-14B
finetune
mit
us
Views
No views yet
Model card
Files and Versions
Community
API
Deploy
Kanitakorn v12 — Thai Education LoRA Adapter
LoRA adapter trained on 6,209 Thai education records targeting P1-M6 (grades 1-12) curriculum:
1,368 real TGAT/A-Level/O-NET exam Q&A (×8 repeat from 171 source records)
1,241 P1-M6 grade×subject synth (20 grade×subject combos, codex-generated)
1,500 v9 Thai-CoT distill
800 OpenThaiEval-style
400 math preservation + 300 AIME math + 300 LCB code + 300 HotpotQA
Recipe: bf16 native LoRA r=64 α=128, LR=3e-5, 2 epochs, gradient_checkpointing, A100-80GB.
Continual on top of
v10
.
train_loss=0.594 (final ~0.40).
Use at inference
Important
: this adapter is trained continual on v10 — for best results load v10 first, then v12 stacked, OR use the merged version.
