Tiny Aya Global, fine-tuned for Indian bank/UPI SMS โ†’ JSON

LoRA adapter for CohereLabs/tiny-aya-global trained with AlignTune on data curated with CuratorKIT and evaluated with AuditKIT. Input: one SMS in English, Hindi or Hinglish. Output: one JSON object with 11 fields (see the dataset card).

Split Exact match Field accuracy Hallucination โ†“ Non-txn F1
test_id 68.7 90.9 10.6 83.3
test_ood 76.7 94.0 1.4 95.5
challenge 59.1 87.3 12.1 78.6

Training path: aligntune ยท prompt format: chat ยท 1600 curated examples ยท LoRA r=16, ฮฑ=32, lr=0.0002, 1 epoch(s).

Licence: the base model is CC-BY-NC (non-commercial). Commercial deployment needs a licence from Cohere. Data: fully synthetic; bank names are fictional; no real customer data.

Downloads last month
94
GGUF
Model size
3B params
Architecture
cohere2
Hardware compatibility
Log In to add your hardware

8-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for Meowniece/tiny-aya-sms-extract-gguf

Adapter
(395)
this model

Collection including Meowniece/tiny-aya-sms-extract-gguf