datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
arena-prose-100-49-models
Arena Prose: 100 prompts × 50 models
A paired exploratory AI-text-detection corpus: 5,000 successful generated responses from 50 models, each answering the same 100 English prose prompts. Generation was performed through OpenRouter in September 2026 with optional reasoning disabled and mandatory reasoning set to low. This is an independent local benchmark inspired by Pangram 4 §5.2, not an official Pangram dataset or exact replication.
Loading
from datasets import… See the full description on the dataset page: https://huggingface.co/datasets/woog/arena-prose-100-49-models.sft-bm-prose
khursanirevo/sft-bm-prose
Bahasa Melayu prose-format text (long-form lessons + textbook-style, ~232k rows).
Splits
split
rows
train
221,170
validation
11,635
Stratified 95/5 by source/category (seed=42).
Source files
data/midtrain/synth_hf_prose.jsonl
data/midtrain/synth_bm_50m.jsonl
Schema
Each row is a JSON object. See the loader script for field details.
Provenance
Generated as part of MaLLaM 2026… See the full description on the dataset page: https://huggingface.co/datasets/khursanirevo/sft-bm-prose.