devim
Datasets
All datasets matching “devim”devim-grounded-turkish
DEVİM Grounded Turkish
This is a methodology and public-evidence repository, not a release of the underlying rights-restricted Turkish corpus.
DEVİM's grounded-data program was created to reduce shortcut learning and weak transfer by linking supervision to source evidence, preserving provenance, separating training material from sequestered evaluation material, and explicitly testing abstention when an answer is not supported.
Verified source frame
Authorized… See the full description on the dataset page: https://huggingface.co/datasets/bazobehram/devim-grounded-turkish.devim-eval
DEVİM Eval
This repository documents DEVİM's evaluation methodology and evaluator failures. It does not release hidden holdout items.
A central research lesson in DEVİM is that evaluator quality is part of the scientific problem: a model can improve language modeling while a capability instrument stays flat, and a seemingly useful QA task can turn out to be solvable by copying.
Frozen V1.8 evaluator
Public-safe facts:
Evaluator: DEVIM_TR_CAPABILITY_V1_8… See the full description on the dataset page: https://huggingface.co/datasets/bazobehram/devim-eval.Devi_Mahatmyacrane-pretokenized-2.4m