--- dataset_info: features: - name: messages list: - name: content dtype: string - name: role dtype: string - name: source dtype: string splits: - name: train num_bytes: 12901675305.337444 num_examples: 3552028 download_size: 5483737746 dataset_size: 12901675305.337444 configs: - config_name: default data_files: - split: train path: data/train-* --- Instruct Data From various Sources: | source | num_samples | language | remarks | |------------------------------------------------------------|-------------|-----------------|--------------------------------------------------| | BAAI/IndustryInstruction | 897196 | en | Filtered Chinese | | BhabhaAI/openhermes-2.5-hindi | 620211 | hi | null | | smol-magpie-ultra-short | 270838 | en | null | | dnyanesh/HindiMathQuest | 250117 | hi | null | | mlabonne/opc-sft-stage2-chat | 224539 | en | null | | manishiitg/hieunguyenminh-roleplay | 159204 | hi | null | | manishiitg/HuggingFaceH4-ultrachat_200k | 147354 | hi | null | | damerajee/camel-phy-chem-bio-sharegpt | 120000 | hi-en mix | null | | DKYoon/dolphin-r1-deepseek-filtered-short | 103423 | en | Filtered Chinese + Small CoT | | sarvamai/samvaad-hi-v1 | 101424 | hi | null | | data/CoT/aqua_rat.json | 59436 | en | null | | mlabonne/lmsys-arena-human-sft-55k | 57362 | en | null | | nothingiisreal/Kalomaze-Opus-Instruct-25k-filtered | 48710 | en | Made BiBo-Specific | | self-oss-instruct | 48071 | en | null | | openhermes-50k | 47492 | en | null | | smol-contraints | 34433 | en | null | | data/CoT/math50k_camel.json | 32870 | en | null | | allenai/tulu-3-sft-personas-instruction-following | 29980 | en | null | | fhai50032/Hindi-Instruct-HQ | 27658 | hi | Filtered Small and Hallucinated Response | | smollm-rewrite-30k | 26657 | en | null | | SmallDoge/SmallThoughts | 26000 | en | Smoll CoT | | smol-summarize-20k | 19272 | en | null | | data/CoT/gsm_rft.json | 18780 | en | null | | data/PoT/mathqa.json | 16344 | en | null | | aaditya/databricks-dolly-15k-Hinglish-Codemix | 15008 | en/hi | null | | fhai50032/medmcqa-solved-thinking-o1 | 14248 | en | CoT | | one-thing/chatbot_arena_conversations_hinglish | 11331 | en/hi | null | | BhabhaAI/Hi-Instruct-v0 | 9969 | hi | null | | HuggingFaceH4/deita-10k-v0-sft | 9959 | en | null | | data/PoT/gsm_gpt4.json | 9697 | en | null | | fhai50032/Hinglish-Preference-Humanized-DPO | 9434 | en/hi | english question - chosen hinglish answer | | fhai50032/Hinglish-Preference-Humanized-DPO ) | 9434 | en/hi | hinglish question - chosen hinglish answer | | data/PoT/numglue.json | 8915 | en | null | | data/CoT/MATH_train.json | 7538 | en | null | | data/PoT/MATH_train.json | 7063 | en | null | | fhai50032/Hinglish-Preference-Humanized-DPO | 6897 | en/hi | hinglish question - rejected hinglish answer | | fhai50032/Hinglish-Preference-Humanized-DPO | 6897 | en/hi | english question - rejected hinglish answer | | data/PoT/aqua_rat_filtered.json | 6436 | en | null | | data/CoT/gsm_train.json | 5012 | en | null | | smol-summarize-5k | 4749 | en | null | | longalign | 3560 | en | null | | macadeliccc/opus_samantha | 3187 | en | Specificed To BiBo | | explore-instruct-rewrite | 3017 | en | null | | HuggingFaceTB/everyday-conversations-llama3.1-2k | 2379 | en | null | | everyday-conversations | 2252 | en | null | | digitalpipelines/samantha-1.1-uncensored | 2038 | en | null | | data/CoT/college_math.json | 1212 | en | null | | OdiaGenAI/roleplay_hindi | 1029 | hi | null | | smangrul/hinglish_self_instruct_v0 | 1018 | en/hi | null | | Abhaykoul/Ancient-Indian-Wisdom | 616 | en | null | | data/PoT/TheoremQA.json | 457 | en | null | | data/CoT/TheoremQA.json | 408 | en | null | | GAIR/o1-journey | 327 | en | null | | data/CoT/number_comparison.json | 202 | en | null | | fhai50032/Aryabhatt-Hinglish | 184 | en/hi | CoT | | fhai50032/Aryabhatt-Hinglish | 184 | en | CoT | ## Language Metrics From the table above (56 rows total), we have three language categories: - **Pure English ("en")**: 39 rows * Total samples ≈ 2,054,859 - **Pure Hindi ("hi")**: 8 rows * Total samples ≈ 1,316,966 - **Mixed (e.g. "en/hi", "hi-en mix")**: 9 rows * Total samples ≈ 180,203 Overall total samples ≈ 3,552,028 **Percentages:** - Pure English: ~57.9% - Pure Hindi: ~37.1% - Mixed: ~5.1%