Kai9987kai/supermix-conversation-datasets

--- license: other task_categories: - text-generation language: - en pretty_name: Supermix Conversation Datasets size_categories: - 100K<n<1M --- # Supermix Conversation Datasets This dataset repo packages the share-ready conversation-training files that were prepared in the Supermix workspace for portable training runs. ## Files - `conversation_data.coding_knowledge_2026_02_19.jsonl` - `conversation_data.mega_creative_250k_v2.jsonl` - `conversation_data.mega_reasoning_creative_v25_75582.jsonl` - `conversation_data.quality_anchor_v2.jsonl` - `conversation_data.supermix_plus_v27_500k.jsonl` - `conversation_data.world_events_2026_02_19.jsonl` ## Format Each file is JSONL. Rows are conversation/training examples intended for Supermix training pipelines. ## Notes - This repo is the curated public bundle from `output/kaggle-upload/supermix-datasets`. - It does not include every intermediate or experimental dataset file from the local workspace. - Check each file's provenance and intended use before mixing it into downstream training.