rounded-off-tiny-aya-bank-sft

Tiny Aya Global fine-tuned as an English + Hindi bank support assistant (benign data only).

Part of Safe on the benchmark, open for fraud (Lexsi Labs AI Safety & Alignment Hackathon, Nov 2026). Code and full results: https://github.com/tayyab415/rounded-off

How it was made: SafeTune PlainSFTTrainer, LoRA r16 merged, lr 1e-4, 2 epochs, 2,985 Banking77 questions with Adaption answers (1,994 EN, 998 hi-IN), cleaned by CuratorKIT.

Base model and license: derived from CohereLabs/tiny-aya-global (CC-BY-NC-4.0); this checkpoint carries the same non-commercial license. Built with Lexsi Labs' SafeTune (Lexsi Source Available License): this model is not for deployment (gate result below).

Intended use: research on safety drift, repair and quantization. Not a production banking assistant.

Bank release gate (article section 7.7): FAIL. It helps with fraud on 15.2% (English) / 13.2% (Hindi) of the bank red-team prompts; the gate passes at 5% or less.

Downloads last month
47
Safetensors
Model size
3B params
Tensor type
BF16
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for tyb343/rounded-off-tiny-aya-bank-sft

Finetuned
(171)
this model

Collection including tyb343/rounded-off-tiny-aya-bank-sft