{ "dataset_id": "ELBennett/big-decisions", "schema_version": "3.0.0", "description": "A tokenizer-neutral collection of decision examples in a unified fixed-choice format. NOUL and SCORE decisions are represented as ordinary choices; option text, soft targets, ordinal ranks, source attribution, and split boundaries are retained.", "format": "Apache Parquet with BEV envelope fields state, questions_json, domain, question_types, question_count, plus source/split provenance fields. In v3, questions_json uses criteria as a list of option strings, label as the selected option string, optional label_probs as an aligned list of weights, and optional choice_ranks as an aligned list of integer ranks.", "training_note": "Tokenize state and choice text with the training model's tokenizer.", "row_counts": { "train": 3632283, "dev": 326539, "test": 86931 }, "source_counts": { "train": { "acos": 10777, "agieval": 6374, "amazon_esci": 46864, "anli": 165553, "api_bank": 10855, "arc_challenge": 2147, "arc_easy": 2070, "banking77": 9999, "bbh": 4948, "bev_decision": 422038, "bfcl": 1738, "bigbench": 359928, "boolq": 8678, "bpomp": 21200, "bright": 843, "cladder": 8025, "clinc150": 26015, "contract_nli": 6872, "cruxeval": 720, "dbpedia14": 91879, "domain_decisions": 113972, "executable_policies": 88388, "finentity": 1787, "forecastbench": 4106, "glue_cola": 8503, "glue_mnli": 392484, "glue_mrpc": 3668, "glue_qnli": 104721, "glue_qqp": 363482, "glue_rte": 2488, "glue_sst2": 66973, "glue_stsb": 5699, "glue_wnli": 635, "go_emotions": 394180, "gpqa": 394, "gsm8k": 8660, "habermas": 7263, "hellaswag": 39905, "home_appliance": 600, "hover": 18028, "humicroedit": 8325, "isarcasm": 4705, "known_posterior": 110550, "logiqa": 14212, "mmlu_full": 9526, "mmlu_high_school": 3070, "mmlu_pro": 10721, "multinli": 220760, "musr": 749, "new_yorker": 10543, "nli4ct": 6629, "phishnchips": 1912, "pop909_cl": 21600, "race": 87911, "ragtruth": 17453, "reclor": 4640, "routerbench": 36141, "sat_analogies": 337, "sata_bench": 1445, "sg_axb": 990, "sg_axg": 320, "sg_boolq": 9427, "sg_cb": 250, "sg_copa": 400, "sg_multirc": 27185, "sg_record": 100658, "sg_rte": 70, "sg_wic": 5428, "sg_wsc": 512, "sgd": 6847, "toolret": 6716, "vast": 10616, "when2call": 17784, "winogrande": 40392 }, "dev": { "acos": 1194, "agieval": 708, "amazon_esci": 794, "anli": 3519, "api_bank": 53, "arc_challenge": 391, "arc_easy": 181, "banking77": 3072, "bbh": 551, "bfcl": 168, "bigbench": 91635, "boolq": 749, "bpomp": 2332, "bright": 94, "cladder": 892, "clinc150": 2251, "contract_nli": 965, "cruxeval": 80, "dbpedia14": 8109, "domain_decisions": 8466, "executable_policies": 7930, "finentity": 198, "forecastbench": 456, "glue_cola": 1034, "glue_mnli": 19645, "glue_mrpc": 408, "glue_qnli": 5463, "glue_qqp": 40422, "glue_rte": 277, "glue_sst2": 872, "glue_stsb": 1498, "glue_wnli": 71, "go_emotions": 34712, "gpqa": 45, "gsm8k": 132, "habermas": 447, "hellaswag": 10042, "home_appliance": 251, "hover": 3977, "isarcasm": 139, "known_posterior": 9450, "logiqa": 1282, "mmlu_full": 2430, "mmlu_high_school": 780, "mmlu_pro": 1251, "multinli": 19232, "musr": 79, "new_yorker": 1131, "nli4ct": 747, "phishnchips": 101, "pop909_cl": 2400, "race": 4888, "ragtruth": 269, "reclor": 483, "routerbench": 342, "sat_analogies": 37, "sata_bench": 160, "sg_axb": 110, "sg_axg": 36, "sg_boolq": 3270, "sg_cb": 56, "sg_copa": 100, "sg_multirc": 4843, "sg_record": 9998, "sg_rte": 6, "sg_wic": 638, "sg_wsc": 94, "sgd": 4043, "toolret": 744, "vast": 1432, "when2call": 617, "winogrande": 1267 }, "test": { "arc_challenge": 1172, "bev_decision": 79889, "logiqa": 935, "race": 4935 } }, "sources": [ { "id": "bev_decision", "title": "BEV Decision Mix", "group": "Decision-making", "origin": "avbiswas/bev-decision", "covers": "Bundled multi-domain choice, yes/no, and ordinal decisions.", "notes": "", "availability": "included", "examples_by_split": { "train": 422038, "test": 79889 }, "examples": 501927, "packed": true, "prepared_examples": 502570 }, { "id": "domain_decisions", "title": "Domain decisions", "group": "Policy", "origin": "altslate/certo-decisions-v2:domain_generator", "covers": "Synthetic domain policy / approval decisions.", "notes": "", "availability": "included", "examples_by_split": { "train": 113972, "dev": 8466 }, "examples": 122438, "packed": true, "prepared_examples": 500000 }, { "id": "go_emotions", "title": "GoEmotions", "group": "Emotion classification", "origin": "google-research-datasets/go_emotions", "covers": "Fine-grained emotion classification over Reddit comments.", "notes": "", "availability": "included", "examples_by_split": { "train": 394180, "dev": 34712 }, "examples": 428892, "packed": true, "prepared_examples": 430001 }, { "id": "executable_policies", "title": "Executable policies", "group": "Policy", "origin": "altslate/certo-decisions-v2:own_policy_generator", "covers": "Synthetic executable policy compliance decisions.", "notes": "", "availability": "included", "examples_by_split": { "train": 88388, "dev": 7930 }, "examples": 96318, "packed": true, "prepared_examples": 192016 }, { "id": "multinli", "title": "MultiNLI", "group": "Semantics and inference", "origin": "nyu-mll/multi_nli", "covers": "Three-way textual entailment (premise/hypothesis).", "notes": "Related to index ANLI, but MultiNLI is the Certo source.", "availability": "included", "examples_by_split": { "train": 220760, "dev": 19232 }, "examples": 239992, "packed": true, "prepared_examples": 120000 }, { "id": "dbpedia14", "title": "DBpedia-14", "group": "Topic classification", "origin": "fancyzhx/dbpedia_14", "covers": "14-way topic classification of Wikipedia abstracts.", "notes": "", "availability": "included", "examples_by_split": { "train": 91879, "dev": 8109 }, "examples": 99988, "packed": true, "prepared_examples": 100000 }, { "id": "known_posterior", "title": "Known posterior", "group": "Probability and belief", "origin": "altslate/certo-decisions-v2:genworld", "covers": "Synthetic calibrated posterior / belief-update decisions.", "notes": "", "availability": "included", "examples_by_split": { "train": 110550, "dev": 9450 }, "examples": 120000, "packed": true, "prepared_examples": 60000 }, { "id": "clinc150", "title": "CLINC150+OOS", "group": "Intent classification", "origin": "clinc/clinc_oos", "covers": "150 banking/assistant intents plus out-of-scope.", "notes": "", "availability": "included", "examples_by_split": { "train": 26015, "dev": 2251 }, "examples": 28266, "packed": true, "prepared_examples": 15000 }, { "id": "boolq", "title": "BoolQ", "group": "Reading comprehension", "origin": "google/boolq", "covers": "Yes/no reading comprehension over Wikipedia passages.", "notes": "", "availability": "included", "examples_by_split": { "train": 8678, "dev": 749 }, "examples": 9427, "packed": true, "prepared_examples": 9427 }, { "id": "arc_easy", "title": "ARC-Easy", "group": "Exam", "origin": "allenai/ai2_arc:ARC-Easy", "covers": "Grade-school science multiple choice (easy split).", "notes": "", "availability": "included", "examples_by_split": { "train": 2070, "dev": 181 }, "examples": 2251, "packed": true, "prepared_examples": 2251 }, { "id": "sat_analogies", "title": "SAT analogies", "group": "Exam", "origin": "willwhitney/6.863-final-project", "covers": "Classic A:B::C:D verbal analogies.", "notes": "", "availability": "included", "examples_by_split": { "train": 337, "dev": 37 }, "examples": 374, "packed": true, "prepared_examples": 374 }, { "id": "mmlu_high_school", "title": "MMLU high school", "group": "Exam", "origin": "cais/mmlu", "covers": "cais/mmlu high_school_* subjects only.", "notes": "Index MMLU also covered by suite mmlu_full for remaining subjects.", "availability": "included", "examples_by_split": { "train": 3070, "dev": 780 }, "examples": 3850, "packed": true, "prepared_examples": 3848 }, { "id": "race", "title": "RACE", "group": "Reading comprehension", "origin": "ehovy/race", "covers": "Middle/high-school reading comprehension MC.", "notes": "", "availability": "included", "examples_by_split": { "train": 87911, "dev": 4888, "test": 4935 }, "examples": 97734, "packed": true, "prepared_examples": 97685 }, { "id": "sata_bench", "title": "SATA-Bench", "group": "Multi-step reasoning", "origin": "sata-bench/sata-bench", "covers": "Select-all-that-apply reading questions.", "notes": "", "availability": "included", "examples_by_split": { "train": 1445, "dev": 160 }, "examples": 1605, "packed": true, "prepared_examples": 1604 }, { "id": "mmlu_pro", "title": "MMLU-Pro", "group": "Exam", "origin": "TIGER-Lab/MMLU-Pro", "covers": "Harder multi-option professional/academic MC.", "notes": "", "availability": "included", "examples_by_split": { "train": 10721, "dev": 1251 }, "examples": 11972, "packed": true, "prepared_examples": 12102 }, { "id": "gpqa", "title": "GPQA", "group": "Exam", "origin": "hendrydong/gpqa_main_mc + gpqa_diamond_mc", "covers": "Graduate-level science MC; diamond subset tagged.", "notes": "", "availability": "included", "examples_by_split": { "train": 394, "dev": 45 }, "examples": 439, "packed": true, "prepared_examples": 438 }, { "id": "musr", "title": "MuSR", "group": "Multi-step reasoning", "origin": "TAUR-Lab/MuSR", "covers": "Multi-step soft reasoning over narratives.", "notes": "", "availability": "included", "examples_by_split": { "train": 749, "dev": 79 }, "examples": 828, "packed": true, "prepared_examples": 756 }, { "id": "reclor", "title": "ReClor", "group": "Exam", "origin": "metaeval/reclor", "covers": "Logical reading comprehension (GMAT/LSAT style).", "notes": "", "availability": "included", "examples_by_split": { "train": 4640, "dev": 483 }, "examples": 5123, "packed": true, "prepared_examples": 5138 }, { "id": "logiqa", "title": "LogiQA", "group": "Exam", "origin": "lgw863/LogiQA-dataset", "covers": "Logical reasoning MC in English and Chinese.", "notes": "", "availability": "included", "examples_by_split": { "train": 14212, "dev": 1282, "test": 935 }, "examples": 16429, "packed": true, "prepared_examples": 16420 }, { "id": "agieval", "title": "AGIEval", "group": "Exam", "origin": "hails/agieval-*", "covers": "Human exam suites (SAT, LSAT, Gaokao, …) as MC.", "notes": "", "availability": "included", "examples_by_split": { "train": 6374, "dev": 708 }, "examples": 7082, "packed": true, "prepared_examples": 7072 }, { "id": "arc_challenge", "title": "ARC-Challenge", "group": "Exam", "origin": "allenai/ai2_arc:ARC-Challenge", "covers": "Harder grade-school science MC; official test held out.", "notes": "", "availability": "included", "examples_by_split": { "train": 2147, "dev": 391, "test": 1172 }, "examples": 3710, "packed": true, "prepared_examples": 3708 }, { "id": "bbh", "title": "BBH", "group": "Reasoning benchmark", "origin": "lukaemon/bbh", "covers": "BIG-Bench Hard multiple-choice / binary tasks.", "notes": "", "availability": "included", "examples_by_split": { "train": 4948, "dev": 551 }, "examples": 5499, "packed": true, "prepared_examples": 5251 }, { "id": "bigbench", "title": "BIG-Bench", "group": "Reasoning benchmark", "origin": "tasksource/bigbench", "covers": "tasksource/bigbench multiple-choice configs.", "notes": "", "availability": "included", "examples_by_split": { "train": 359928, "dev": 91635 }, "examples": 451563, "packed": true, "prepared_examples": 901085 }, { "id": "banking77", "title": "BANKING77", "group": "Intent classification", "origin": "PolyAI/banking77", "covers": "77-way banking customer-intent classification.", "notes": "", "availability": "included", "examples_by_split": { "train": 9999, "dev": 3072 }, "examples": 13071, "packed": true, "prepared_examples": 13083 }, { "id": "gsm8k", "title": "GSM8K", "group": "Mathematical reasoning", "origin": "openai/gsm8k", "covers": "Grade-school math word problems as 4-way MC.", "notes": "Numeric distractors are generated by mutating the gold answer.", "availability": "included", "examples_by_split": { "train": 8660, "dev": 132 }, "examples": 8792, "packed": true, "prepared_examples": 8792 }, { "id": "winogrande", "title": "WinoGrande", "group": "Commonsense reasoning", "origin": "allenai/winogrande", "covers": "Commonsense pronoun / gap fill (2 options).", "notes": "", "availability": "included", "examples_by_split": { "train": 40392, "dev": 1267 }, "examples": 41659, "packed": true, "prepared_examples": 41659 }, { "id": "hellaswag", "title": "HellaSwag", "group": "Commonsense reasoning", "origin": "Rowan/hellaswag", "covers": "Commonsense sentence completion (4 endings).", "notes": "", "availability": "included", "examples_by_split": { "train": 39905, "dev": 10042 }, "examples": 49947, "packed": true, "prepared_examples": 49947 }, { "id": "anli", "title": "ANLI", "group": "Semantics and inference", "origin": "facebook/anli", "covers": "Adversarial three-way NLI (rounds 1–3).", "notes": "", "availability": "included", "examples_by_split": { "train": 165553, "dev": 3519 }, "examples": 169072, "packed": true, "prepared_examples": 169072 }, { "id": "contract_nli", "title": "ContractNLI", "group": "Semantics and inference", "origin": "tasksource/contract-nli:contractnli_a", "covers": "NDA hypothesis support / contradict / not mentioned.", "notes": "", "availability": "included", "examples_by_split": { "train": 6872, "dev": 965 }, "examples": 7837, "packed": true, "prepared_examples": 7837 }, { "id": "mmlu_full", "title": "MMLU (remaining)", "group": "Exam", "origin": "cais/mmlu", "covers": "All cais/mmlu subjects except high_school_* (already in capacity).", "notes": "", "availability": "included", "examples_by_split": { "train": 9526, "dev": 2430 }, "examples": 11956, "packed": true, "prepared_examples": 11967 }, { "id": "cladder", "title": "CLadder", "group": "Causal reasoning", "origin": "causalNLP/cladder (HF mirror or GitHub release)", "covers": "Yes/no causal reasoning over stated worlds.", "notes": "", "availability": "included", "examples_by_split": { "train": 8025, "dev": 892 }, "examples": 8917, "packed": true, "prepared_examples": 8917 }, { "id": "cruxeval", "title": "CRUXEval", "group": "Code reasoning", "origin": "cruxeval-org/cruxeval", "covers": "Predict Python function output from input (MC).", "notes": "Distractors sampled from other problems' outputs.", "availability": "included", "examples_by_split": { "train": 720, "dev": 80 }, "examples": 800, "packed": true, "prepared_examples": 800 }, { "id": "hle", "title": "HLE", "group": "Exam", "origin": "cais/hle", "covers": "Humanity's Last Exam text-only original MC items.", "notes": "Explicitly excluded from this published dataset by user instruction.", "availability": "omitted_by_user", "examples_by_split": {}, "examples": 0, "packed": false, "prepared_examples": 0 }, { "id": "chessbench", "title": "ChessBench", "group": "Game reasoning", "origin": "storage.googleapis.com/searchless_chess (action_value bags)", "covers": "Best legal move among candidates for a FEN.", "notes": "Explicitly excluded from this published dataset by user instruction.", "availability": "omitted_by_user", "examples_by_split": {}, "examples": 0, "packed": false, "prepared_examples": 0 }, { "id": "forecastbench", "title": "ForecastBench", "group": "Forecasting", "origin": "forecastingresearch/forecastbench-datasets", "covers": "Binary event forecasts from frozen pre-resolution context.", "notes": "Local: data/suite-raw/forecastbench/{question_sets,resolution_sets}. Join resolved ids; pack yes/no (Brier is the index metric).", "availability": "included", "examples_by_split": { "train": 4106, "dev": 456 }, "examples": 4562, "packed": true, "prepared_examples": 4562 }, { "id": "finentity", "title": "FinEntity", "group": "Financial sentiment", "origin": "yixuantt/FinEntity", "covers": "Entity-conditioned financial sentiment (pos/neg/neu).", "notes": "", "availability": "included", "examples_by_split": { "train": 1787, "dev": 198 }, "examples": 1985, "packed": true, "prepared_examples": 1985 }, { "id": "isarcasm", "title": "iSarcasmEval", "group": "Sarcasm detection", "origin": "iabufarha/iSarcasmEval", "covers": "English sarcasm detection (track A).", "notes": "", "availability": "included", "examples_by_split": { "train": 4705, "dev": 139 }, "examples": 4844, "packed": true, "prepared_examples": 4844 }, { "id": "vast", "title": "VAST", "group": "Stance classification", "origin": "emilyallaway/zero-shot-stance", "covers": "Zero-shot stance: favor / against / neutral on a topic.", "notes": "", "availability": "included", "examples_by_split": { "train": 10616, "dev": 1432 }, "examples": 12048, "packed": true, "prepared_examples": 12048 }, { "id": "nli4ct", "title": "NLI4CT", "group": "Semantics and inference", "origin": "ai-systems/Task-2-SemEval-2024", "covers": "Clinical-trial statement support / contradiction.", "notes": "", "availability": "included", "examples_by_split": { "train": 6629, "dev": 747 }, "examples": 7376, "packed": true, "prepared_examples": 7376 }, { "id": "acos", "title": "ACOS", "group": "Aspect and sentiment classification", "origin": "NUSTM/ACOS", "covers": "Aspect-category × sentiment presence in reviews.", "notes": "", "availability": "included", "examples_by_split": { "train": 10777, "dev": 1194 }, "examples": 11971, "packed": true, "prepared_examples": 11971 }, { "id": "amazon_esci", "title": "Amazon ESCI", "group": "Search relevance", "origin": "tasksource/esci or amazon-science/esci-data", "covers": "Query–product relevance: Exact/Substitute/Complement/Irrelevant.", "notes": "", "availability": "included", "examples_by_split": { "train": 46864, "dev": 794 }, "examples": 47658, "packed": true, "prepared_examples": 47658 }, { "id": "ragtruth", "title": "RAGTruth", "group": "Factuality detection", "origin": "wandb/RAGTruth-processed", "covers": "Detect hallucinated spans in RAG answers.", "notes": "", "availability": "included", "examples_by_split": { "train": 17453, "dev": 269 }, "examples": 17722, "packed": true, "prepared_examples": 17722 }, { "id": "bright", "title": "BRIGHT", "group": "Search relevance", "origin": "xlangai/BRIGHT", "covers": "Hard reasoning retrieval: pick helpful doc among candidates.", "notes": "examples+documents; gold doc vs sampled negatives (skips huge domains).", "availability": "included", "examples_by_split": { "train": 843, "dev": 94 }, "examples": 937, "packed": true, "prepared_examples": 937 }, { "id": "hover", "title": "HoVer", "group": "Fact verification", "origin": "vincentkoc/hover-parquet", "covers": "Multi-hop claim verification (supported / not).", "notes": "", "availability": "included", "examples_by_split": { "train": 18028, "dev": 3977 }, "examples": 22005, "packed": true, "prepared_examples": 22005 }, { "id": "sgd", "title": "SGD/SGD-X", "group": "Dialogue intent classification", "origin": "google-research-datasets/dstc8-schema-guided-dialogue", "covers": "Schema-guided dialogue current-turn intent.", "notes": "", "availability": "included", "examples_by_split": { "train": 6847, "dev": 4043 }, "examples": 10890, "packed": true, "prepared_examples": 10950 }, { "id": "phishnchips", "title": "PhishNChips", "group": "Security classification", "origin": "AreLit/PhishNChips", "covers": "Phishing vs legitimate email/link decisions.", "notes": "HF config `emails`; phish_label 1/0 with email_content JSON.", "availability": "included", "examples_by_split": { "train": 1912, "dev": 101 }, "examples": 2013, "packed": true, "prepared_examples": 2013 }, { "id": "bfcl", "title": "BFCL", "group": "Tool use", "origin": "gorilla-llm/gorilla berkeley-function-call-leaderboard", "covers": "Per-function call / no-call decisions for a request.", "notes": "Local data/suite-raw/bfcl from gorilla BFCL_v4_* + possible_answer.", "availability": "included", "examples_by_split": { "train": 1738, "dev": 168 }, "examples": 1906, "packed": true, "prepared_examples": 2180 }, { "id": "toolret", "title": "ToolRet", "group": "Tool use", "origin": "mangopy/ToolRet-Queries", "covers": "Pick the most useful tool for a request among candidates.", "notes": "All ToolRet-Queries configs; gold = first relevance>0 label.", "availability": "included", "examples_by_split": { "train": 6716, "dev": 744 }, "examples": 7460, "packed": true, "prepared_examples": 7960 }, { "id": "api_bank", "title": "API-Bank", "group": "Tool use", "origin": "liminghao1630/API-Bank", "covers": "Next API to call from a fixed catalog given dialogue.", "notes": "Local data/suite-raw/api_bank lv*-api-train + level-*-api JSON.", "availability": "included", "examples_by_split": { "train": 10855, "dev": 53 }, "examples": 10908, "packed": true, "prepared_examples": 10908 }, { "id": "routerbench", "title": "RouterBench", "group": "Model routing", "origin": "withmartian/routerbench", "covers": "Route a prompt to the best of several candidate models.", "notes": "", "availability": "included", "examples_by_split": { "train": 36141, "dev": 342 }, "examples": 36483, "packed": true, "prepared_examples": 72994 }, { "id": "when2call", "title": "When2Call", "group": "Tool use", "origin": "nvidia/When2Call", "covers": "When to call a tool / ask / abstain given tool specs.", "notes": "", "availability": "included", "examples_by_split": { "train": 17784, "dev": 617 }, "examples": 18401, "packed": true, "prepared_examples": 18401 }, { "id": "home_appliance", "title": "Home appliance simulator", "group": "Tool use", "origin": "apolinario/decision-index home_appliance.py", "covers": "Household command handling checklist decisions.", "notes": "Generated cases in data/suite-raw/home_appliance/cases.jsonl.", "availability": "included", "examples_by_split": { "train": 600, "dev": 251 }, "examples": 851, "packed": true, "prepared_examples": 2726 }, { "id": "bpomp", "title": "BPoMP", "group": "Humor and language", "origin": "zenodo.org/records/7299879", "covers": "Original vs meter/rhyme/sense-broken limerick.", "notes": "Local BPoMP_datasets_p*_out_of_3.json under data/suite-raw/bpomp/.", "availability": "included", "examples_by_split": { "train": 21200, "dev": 2332 }, "examples": 23532, "packed": true, "prepared_examples": 23757 }, { "id": "humicroedit", "title": "Humicroedit", "group": "Humor and language", "origin": "cs.rochester.edu humicroedit SemEval-2020 task-7", "covers": "Which one-word headline edit humans rated funnier.", "notes": "Local data/suite-raw/humicroedit/data/task-2/train.csv.", "availability": "included", "examples_by_split": { "train": 8325 }, "examples": 8325, "packed": true, "prepared_examples": 8353 }, { "id": "pop909_cl", "title": "POP909-CL", "group": "Music understanding", "origin": "AndyWeasley2004/POP909-CL-Dataset", "covers": "Symbolic pop chord recognition at a beat (sampled options).", "notes": "Local MIDI under data/suite-raw/pop909/POP909_processed/. Chord track → root:quality labels; catalog sampled to ≤8 options.", "availability": "included", "examples_by_split": { "train": 21600, "dev": 2400 }, "examples": 24000, "packed": true, "prepared_examples": 24000 }, { "id": "habermas", "title": "Habermas Machine", "group": "Group preference", "origin": "google-deepmind/habermas_machine", "covers": "Which candidate statement a group ranked highest.", "notes": "Local GCS parquet data/suite-raw/habermas/hm_all_candidate_comparisons.parquet; rank 0 = best.", "availability": "included", "examples_by_split": { "train": 7263, "dev": 447 }, "examples": 7710, "packed": true, "prepared_examples": 8084 }, { "id": "new_yorker", "title": "New Yorker", "group": "Humor and language", "origin": "jmhessel/newyorker_caption_contest", "covers": "Caption matching / ranking for New Yorker cartoons (text).", "notes": "Text-only matching/ranking configs; image-from-pixels skipped.", "availability": "included", "examples_by_split": { "train": 10543, "dev": 1131 }, "examples": 11674, "packed": true, "prepared_examples": 11723 }, { "id": "glue_cola", "title": "GLUE CoLA", "group": "Grammar and acceptability", "origin": "nyu-mll/glue:cola", "covers": "Acceptability judgment for an English sentence.", "notes": "", "availability": "included", "examples_by_split": { "train": 8503, "dev": 1034 }, "examples": 9537, "packed": true, "prepared_examples": 9537 }, { "id": "glue_sst2", "title": "GLUE SST-2", "group": "Sentiment classification", "origin": "nyu-mll/glue:sst2", "covers": "Binary sentiment for a sentence.", "notes": "", "availability": "included", "examples_by_split": { "train": 66973, "dev": 872 }, "examples": 67845, "packed": true, "prepared_examples": 67845 }, { "id": "glue_mrpc", "title": "GLUE MRPC", "group": "Semantics and inference", "origin": "nyu-mll/glue:mrpc", "covers": "Paraphrase / semantic equivalence of two sentences.", "notes": "", "availability": "included", "examples_by_split": { "train": 3668, "dev": 408 }, "examples": 4076, "packed": true, "prepared_examples": 4076 }, { "id": "glue_stsb", "title": "GLUE STS-B", "group": "Semantic similarity", "origin": "nyu-mll/glue:stsb", "covers": "Semantic similarity binned to integer scores 0–5.", "notes": "", "availability": "included", "examples_by_split": { "train": 5699, "dev": 1498 }, "examples": 7197, "packed": true, "prepared_examples": 7197 }, { "id": "glue_qqp", "title": "GLUE QQP", "group": "Semantic similarity", "origin": "nyu-mll/glue:qqp", "covers": "Duplicate-question detection.", "notes": "", "availability": "included", "examples_by_split": { "train": 363482, "dev": 40422 }, "examples": 403904, "packed": true, "prepared_examples": 403904 }, { "id": "glue_mnli", "title": "GLUE MNLI", "group": "Semantics and inference", "origin": "nyu-mll/glue:mnli", "covers": "Three-way NLI (entailment / neutral / contradiction).", "notes": "Overlaps Certo multinli family.", "availability": "included", "examples_by_split": { "train": 392484, "dev": 19645 }, "examples": 412129, "packed": true, "prepared_examples": 412129 }, { "id": "glue_qnli", "title": "GLUE QNLI", "group": "Semantics and inference", "origin": "nyu-mll/glue:qnli", "covers": "Does a sentence answer a question (entailment pair).", "notes": "", "availability": "included", "examples_by_split": { "train": 104721, "dev": 5463 }, "examples": 110184, "packed": true, "prepared_examples": 110184 }, { "id": "glue_rte", "title": "GLUE RTE", "group": "Semantics and inference", "origin": "nyu-mll/glue:rte", "covers": "Binary textual entailment.", "notes": "", "availability": "included", "examples_by_split": { "train": 2488, "dev": 277 }, "examples": 2765, "packed": true, "prepared_examples": 2765 }, { "id": "glue_wnli", "title": "GLUE WNLI", "group": "Semantics and inference", "origin": "nyu-mll/glue:wnli", "covers": "Winograd-style entailment pairs.", "notes": "", "availability": "included", "examples_by_split": { "train": 635, "dev": 71 }, "examples": 706, "packed": true, "prepared_examples": 706 }, { "id": "sg_boolq", "title": "SuperGLUE BoolQ", "group": "Reading comprehension", "origin": "aps/super_glue:boolq", "covers": "Yes/no reading comprehension over a passage.", "notes": "Overlaps Certo boolq; kept for SuperGLUE completeness.", "availability": "included", "examples_by_split": { "train": 9427, "dev": 3270 }, "examples": 12697, "packed": true, "prepared_examples": 12697 }, { "id": "sg_cb", "title": "SuperGLUE CommitmentBank", "group": "Semantics and inference", "origin": "aps/super_glue:cb", "covers": "Three-way commitment / NLI over short premises.", "notes": "", "availability": "included", "examples_by_split": { "train": 250, "dev": 56 }, "examples": 306, "packed": true, "prepared_examples": 306 }, { "id": "sg_copa", "title": "SuperGLUE COPA", "group": "Causal reasoning", "origin": "aps/super_glue:copa", "covers": "Choose the more plausible cause or effect.", "notes": "", "availability": "included", "examples_by_split": { "train": 400, "dev": 100 }, "examples": 500, "packed": true, "prepared_examples": 500 }, { "id": "sg_multirc", "title": "SuperGLUE MultiRC", "group": "Reading comprehension", "origin": "aps/super_glue:multirc", "covers": "Is a candidate answer correct for a multi-sentence question?", "notes": "", "availability": "included", "examples_by_split": { "train": 27185, "dev": 4843 }, "examples": 32028, "packed": true, "prepared_examples": 32028 }, { "id": "sg_record", "title": "SuperGLUE ReCoRD", "group": "Reading comprehension", "origin": "aps/super_glue:record", "covers": "Cloze entity fill from a candidate entity set.", "notes": "Options sampled from the entity set when it is larger than 8.", "availability": "included", "examples_by_split": { "train": 100658, "dev": 9998 }, "examples": 110656, "packed": true, "prepared_examples": 110730 }, { "id": "sg_rte", "title": "SuperGLUE RTE", "group": "Semantics and inference", "origin": "aps/super_glue:rte", "covers": "Binary textual entailment (SuperGLUE split).", "notes": "", "availability": "included", "examples_by_split": { "train": 70, "dev": 6 }, "examples": 76, "packed": true, "prepared_examples": 76 }, { "id": "sg_wic", "title": "SuperGLUE WiC", "group": "Lexical semantics", "origin": "aps/super_glue:wic", "covers": "Word-in-context same-sense decision.", "notes": "", "availability": "included", "examples_by_split": { "train": 5428, "dev": 638 }, "examples": 6066, "packed": true, "prepared_examples": 6066 }, { "id": "sg_wsc", "title": "SuperGLUE WSC", "group": "Coreference resolution", "origin": "aps/super_glue:wsc.fixed", "covers": "Winograd schema coreference (span refers?).", "notes": "", "availability": "included", "examples_by_split": { "train": 512, "dev": 94 }, "examples": 606, "packed": true, "prepared_examples": 606 }, { "id": "sg_axb", "title": "SuperGLUE AX-b", "group": "Semantics and inference", "origin": "aps/super_glue:axb", "covers": "Broad diagnostic entailment pairs (labeled test).", "notes": "Packed as pool from the labeled test diagnostic.", "availability": "included", "examples_by_split": { "train": 990, "dev": 110 }, "examples": 1100, "packed": true, "prepared_examples": 1100 }, { "id": "sg_axg", "title": "SuperGLUE AX-g", "group": "Semantics and inference", "origin": "aps/super_glue:axg", "covers": "Gender diagnostic entailment pairs (labeled test).", "notes": "Packed as pool from the labeled test diagnostic.", "availability": "included", "examples_by_split": { "train": 320, "dev": 36 }, "examples": 356, "packed": true, "prepared_examples": 356 } ], "task_groups": { "bev_decision": "Decision-making", "domain_decisions": "Policy", "go_emotions": "Emotion classification", "executable_policies": "Policy", "multinli": "Semantics and inference", "dbpedia14": "Topic classification", "known_posterior": "Probability and belief", "clinc150": "Intent classification", "boolq": "Reading comprehension", "arc_easy": "Exam", "sat_analogies": "Exam", "mmlu_high_school": "Exam", "race": "Reading comprehension", "sata_bench": "Multi-step reasoning", "mmlu_pro": "Exam", "gpqa": "Exam", "musr": "Multi-step reasoning", "reclor": "Exam", "logiqa": "Exam", "agieval": "Exam", "arc_challenge": "Exam", "bbh": "Reasoning benchmark", "bigbench": "Reasoning benchmark", "banking77": "Intent classification", "gsm8k": "Mathematical reasoning", "winogrande": "Commonsense reasoning", "hellaswag": "Commonsense reasoning", "anli": "Semantics and inference", "contract_nli": "Semantics and inference", "mmlu_full": "Exam", "cladder": "Causal reasoning", "cruxeval": "Code reasoning", "hle": "Exam", "chessbench": "Game reasoning", "forecastbench": "Forecasting", "finentity": "Financial sentiment", "isarcasm": "Sarcasm detection", "vast": "Stance classification", "nli4ct": "Semantics and inference", "acos": "Aspect and sentiment classification", "amazon_esci": "Search relevance", "ragtruth": "Factuality detection", "bright": "Search relevance", "hover": "Fact verification", "sgd": "Dialogue intent classification", "phishnchips": "Security classification", "bfcl": "Tool use", "toolret": "Tool use", "api_bank": "Tool use", "routerbench": "Model routing", "when2call": "Tool use", "home_appliance": "Tool use", "bpomp": "Humor and language", "humicroedit": "Humor and language", "pop909_cl": "Music understanding", "habermas": "Group preference", "new_yorker": "Humor and language", "glue_cola": "Grammar and acceptability", "glue_sst2": "Sentiment classification", "glue_mrpc": "Semantics and inference", "glue_stsb": "Semantic similarity", "glue_qqp": "Semantic similarity", "glue_mnli": "Semantics and inference", "glue_qnli": "Semantics and inference", "glue_rte": "Semantics and inference", "glue_wnli": "Semantics and inference", "sg_boolq": "Reading comprehension", "sg_cb": "Semantics and inference", "sg_copa": "Causal reasoning", "sg_multirc": "Reading comprehension", "sg_record": "Reading comprehension", "sg_rte": "Semantics and inference", "sg_wic": "Lexical semantics", "sg_wsc": "Coreference resolution", "sg_axb": "Semantics and inference", "sg_axg": "Semantics and inference" }, "base_source": { "dataset_id": "ELBennett/big-decisions", "revision": "28acab3392071694a0544edb6ba4e649a2d9c6ed", "published_input": true }, "prepared_stores": { "stores": [ "certo", "capacity", "exam", "suite-verified", "suite-extra", "banking77" ] }, "parquet_counts": { "train": 3632283, "dev": 326539, "test": 86931 }, "datasets_load_validation": { "train": 3632283, "dev": 326539, "test": 86931 }, "bev_source": { "dataset_id": "avbiswas/bev-decision", "revision": "11ed9b5b879e7f9524e7c89c1037d498c54206f0", "config": "all", "source_splits": { "train": "train", "test": "test" }, "input_rows": { "train": 231332, "test": 44856 }, "expanded_questions": { "train": 422551, "test": 80019 }, "retained_decisions": { "train": 422040, "test": 79889 }, "deduplicated_by_source": { "train": { "bev_decision": 511 }, "test": { "bev_decision": 130 } } }, "exclusions": { "selected_out_before_conversion": [ "cfcolor", "chessbench", "hle" ] }, "deduplication": { "semantic_key": "normalized state + instructions + option text + target distribution + ordinal rank", "split_priority": [ "train", "dev", "test" ], "first_eligible_source_attribution_retained": true, "duplicate_decisions_removed_by_split_and_source": { "train": { "dbpedia14": 91897, "multinli": 8, "go_emotions": 396062, "clinc150": 1613, "arc_easy": 2070, "arc_challenge": 2145, "boolq": 8678, "domain_decisions": 806668, "executable_policies": 264556, "race": 87783, "sat_analogies": 337, "mmlu_high_school": 3064, "sata_bench": 1443, "bigbench": 1076440, "reclor": 4636, "logiqa": 14194, "mmlu_pro": 10935, "gpqa": 392, "musr": 611, "agieval": 6358, "bbh": 4500, "humicroedit": 8329, "routerbench": 62400, "home_appliance": 1554, "banking77": 10007, "new_yorker": 41, "sgd": 37, "sg_record": 72, "mmlu_full": 10, "bpomp": 183, "bfcl": 224, "bev_decision": 511 }, "dev": { "dbpedia14": 8115, "go_emotions": 35048, "clinc150": 121, "arc_easy": 181, "arc_challenge": 389, "boolq": 749, "domain_decisions": 70894, "executable_policies": 23158, "mmlu_high_school": 778, "race": 4884, "sat_analogies": 37, "sata_bench": 160, "bigbench": 274113, "mmlu_pro": 1295, "reclor": 517, "logiqa": 1282, "gpqa": 45, "musr": 73, "agieval": 704, "bbh": 501, "home_appliance": 1684, "routerbench": 10608, "banking77": 3088, "mmlu_full": 1, "new_yorker": 8, "sgd": 23, "sg_record": 2, "bpomp": 42, "bfcl": 50 }, "test": { "race": 4931, "logiqa": 935, "arc_challenge": 1172, "bev_decision": 130 } }, "duplicate_rows_removed_by_split_and_source": { "train": { "dbpedia14": 91897, "multinli": 8, "go_emotions": 396062, "clinc150": 1613, "arc_easy": 2070, "arc_challenge": 2145, "boolq": 8678, "domain_decisions": 806668, "executable_policies": 264556, "race": 87783, "sat_analogies": 337, "mmlu_high_school": 3064, "sata_bench": 1443, "bigbench": 1076440, "reclor": 4636, "logiqa": 14194, "mmlu_pro": 10935, "gpqa": 392, "musr": 611, "agieval": 6358, "bbh": 4500, "humicroedit": 8329, "routerbench": 62400, "home_appliance": 1554, "banking77": 10007, "new_yorker": 41, "sgd": 37, "sg_record": 72, "mmlu_full": 10, "bpomp": 183, "bfcl": 224, "bev_decision": 511 }, "dev": { "dbpedia14": 8115, "go_emotions": 35048, "clinc150": 121, "arc_easy": 181, "arc_challenge": 389, "boolq": 749, "domain_decisions": 70894, "executable_policies": 23158, "mmlu_high_school": 778, "race": 4884, "sat_analogies": 37, "sata_bench": 160, "bigbench": 274113, "mmlu_pro": 1295, "reclor": 517, "logiqa": 1282, "gpqa": 45, "musr": 73, "agieval": 704, "bbh": 501, "home_appliance": 1684, "routerbench": 10608, "banking77": 3088, "mmlu_full": 1, "new_yorker": 8, "sgd": 23, "sg_record": 2, "bpomp": 42, "bfcl": 50 }, "test": { "race": 4931, "logiqa": 935, "arc_challenge": 1172, "bev_decision": 130 } }, "cross_split_duplicates_retained": 0 }, "source_row_totals": { "retained_rows": 4045753, "duplicates_removed": 3313476, "duplicate_decisions_removed": 3313476, "duplicate_rows_removed": 3313476, "malformed_base_rows_dropped_by_source": { "bbh": 1, "humicroedit": 26, "mmlu_high_school": 2, "mmlu_pro": 1, "race": 19 }, "malformed_choice_rows_dropped_v3": { "train": { "race": 17, "mmlu_high_school": 1, "mmlu_pro": 1, "bbh": 1, "humicroedit": 26, "habermas": 352, "toolret": 450, "bev_decision": 2 }, "dev": { "race": 1, "mmlu_high_school": 1, "toolret": 50, "habermas": 22 }, "test": { "race": 1 } } }, "note": "HLE, ChessBench, and CFColor were excluded before conversion. Source identity, revision, config, split, and question keys are preserved in provenance.", "schema_conversion": { "from_schema_version": "2.0.0", "question_type": "choice for every question", "criteria": "ordered list of option strings; choice_ranks order first when present, otherwise preserve v2 deterministic order", "label": "zero-based integer index into criteria", "label_probs": "optional list of probabilities aligned with criteria; present only for soft-target examples", "choice_ranks": "optional integer list aligned with criteria; preserves ordinal SCORE order and rank", "questions": { "questions": 4045753, "criterion_items": 15998658, "rows": 4045753, "binary_yes_no_questions": 786569, "soft_target_questions": 129638, "ranked_questions": 69108, "invalid_single_option_questions": 925, "invalid_single_option_rows": 925 }, "split_row_counts_unchanged": false, "split_row_counts": { "train": 3632283, "dev": 326539, "test": 86931 }, "dropped_invalid_single_option_rows": { "train": { "race": 17, "mmlu_high_school": 1, "mmlu_pro": 1, "bbh": 1, "humicroedit": 26, "habermas": 352, "toolret": 450, "bev_decision": 2 }, "dev": { "race": 1, "mmlu_high_school": 1, "toolret": 50, "habermas": 22 }, "test": { "race": 1 } } } }