Datasets:

amyrmahdy
/

decima-kg-judgments

id string	config string	task string	origin string	kind string	state string	question string	choices list	gold int64	probs list
6f5ed390b04c737c	entity_matching	em	procedural	score	{"entity_a": {"title": "dance away", "artist": "Shaipiost & the Sparrows", "album": "Golden Sparrows", "genre": "Folk", "duration": "4m 27s"}, "entity_b": {"title": "Dance Away", "artist": "Shaipiost & the Sparrows", "album": "Golden Sparrows", "genre": "Singer/Songwriter", "duration": "4:27"}}	How likely is it that both records describe one and the same track?	[ "Different tracks.", "Related tracks (a variant, version or edition of each other); possibly the same.", "The same track." ]	2	[ 0.0124, 0.0496, 0.938 ]
5afb981bfa3a5ca6	entity_matching	em	procedural	verify	{"entity_a": {"package": "retry-path", "maintainer": "arjunlaur", "registry": "Maven", "version": "9.20.1 (stable)", "licence": "New BSD"}, "entity_b": {"package": "color-yaml", "maintainer": "Prozean", "registry": "pypi", "version": "version 2.23.1"}}	Are `entity_a` and `entity_b` the same package, not just a related variant? True if: One and the same package. False if: Two different packages, or two variants of one.	[ "yes", "no" ]	1	[ 0.071, 0.929 ]
b2fc8ad1d8731c0e	entity_matching	em	procedural	verify	{"a": {"name": "Stanil XR-30", "maker": "Dunecrest", "product_type": "Cameras", "color": "Sand", "price": "1755.83"}, "b": {"name": "Alderfield Stanil R729", "maker": "Alderfield", "product_type": "Wireless router", "price": "1163.03 USD"}}	Is the price the same in `a` and `b`?	[ "yes", "no" ]	1	[ 0.071, 0.929 ]
6a9bff05447cd2a2	entity_matching	em	procedural	verify	{"record_a": {"name": "CRANE FIELD TERRACES", "winery": "Birchford Wines Ltd", "varietal": "Steen", "year": "2021", "alcohol": "13.0%", "region": "Kettle Valley", "volume": "0.375 L"}, "record_b": {"name": "Field Estate", "winery": "Tealhaven", "varietal": "Cabernet-Sauvignon", "year": "2023", "alcohol": "14.50 %", "vo...	Is the product named the same in both records?	[ "yes", "no" ]	1	[ 0.041, 0.959 ]
c0d21ed81652f77a	entity_matching	em	procedural	verify	{"entity_a": {"name": "R70", "make": "Crodros Motors", "type": "Coupe", "model_year": 2010, "grade": "SE", "displacement": "1.6L"}, "entity_b": {"name": "R70", "make": "Crodros Motors", "type": "Coupe", "model_year": 2010, "grade": "SE", "displacement": "1.6L"}}	Same name? Compare `entity_a.name` with `entity_b.name`.	[ "yes", "no" ]	0	[ 0.951, 0.049 ]
d965c0cb5964bf85	entity_matching	em	procedural	score	{"entity_a": {"song": "Wait Away", "artist_name": "Tastabol & the Voyages", "genre": "EDM", "duration": "3:56 min", "price": "1.29"}, "entity_b": {"song": "Wait Away", "artist_name": "Tastabol & the Voyages", "album_name": "sumer orchards", "genre": "Electronic", "price": "1.29"}}	Are these two track records about the same real-world track?	[ "They refer to two different real-world entities.", "They refer to closely related entities that may or may not be the same one.", "They refer to one and the same real-world entity." ]	2	[ 0.0144, 0.0576, 0.928 ]
952a982ae5931747	entity_matching	em	procedural	score	{"entity_a": {"company": "Birch Grove Bank", "city": "Osaka", "sector": "Banks", "symbol": "NYSE:BOE", "domain": "www.birchgrove.net", "founded_year": "1895"}, "entity_b": {"company": "Zereth Dynamics SA", "city": "Boston", "sector": "Aerospace", "symbol": "SIX: ZER", "domain": "https://zereth.net", "founded_year": "19...	How likely is it that both records describe one and the same company?	[ "They refer to two different real-world entities.", "They refer to closely related entities that may or may not be the same one.", "They refer to one and the same real-world entity." ]	0	[ 0.937, 0.0504, 0.0126 ]
ecb35637a08a493a	entity_matching	em	procedural	verify	{"entity_a": {"name": "SORRELFELL AEROSPACE GMBH", "city": "LA", "industry": "Aerospace & Defense", "symbol": "TSX:SRFL"}, "entity_b": {"name": "sorrelfell aerospace gmbh", "symbol": "TSX: SRFL", "industry": "aerospace & defense", "url": "www.sorrelfell.net", "city": "Los Angeles, CA", "founded_year": "1933"}}	Setting formatting aside, do the two records carry the same name? True if: Same name, possibly written differently. False if: Different names.	[ "yes", "no" ]	0	[ 0.949, 0.051 ]
77135b4324b4382d	entity_matching	em	procedural	score	{"entity_a": {"name": "trethas cafe", "addr": "2151 Harrow Dr.", "city": "München", "telephone": "881/845-6793", "cuisine": "Italian"}, "entity_b": {"name": "The Silent Trethas", "addr": "1369 Calloway Doctor", "city": "Zurich", "telephone": "+1 264 306 2857", "cuisine": "American"}}	Judge whether `entity_a` and `entity_b` are duplicate entries for one restaurant, ignoring formatting differences.	[ "distinct", "related variant", "same" ]	0	[ 0.957, 0.0344, 0.0086 ]
4ae30dcd4a33efcc	entity_matching	em	procedural	score	{"entity_a": {"title": "Run On", "artist_name": "quothefys & the feathers", "genre": "pop", "price": "$1.29"}, "entity_b": {"title": "run on", "artist_name": "Quothefys & The Feathers", "album": "silent shores", "genre": "Pop", "length": "2:10 min", "price": "$ 1.29"}}	Entity resolution: how do `entity_a` and `entity_b` relate?	[ "They describe two different tracks.", "They describe closely related tracks that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same track." ]	2	[ 0.0084, 0.0336, 0.958 ]
3fc12edf985eb4ff	entity_matching	em	procedural	verify	{"entity_a": {"name": "Brecenav Kitchen", "street": "18 Fairmount Drive", "city": "Lyon, France", "phone": "(707) 558-6070", "cuisine": "Korean BBQ"}, "entity_b": {"name": "Breceav Kitchen", "street": "18 Fairmount Drive", "city": "Lyon, France", "phone": "+1 707 558 6070", "cuisine": "Korean"}}	Do `entity_a.cuisine` and `entity_b.cuisine` name the same cuisine?	[ "yes", "no" ]	0	[ 0.946, 0.054 ]
0c7d1ace0e85cce4	entity_matching	em	procedural	verify	{"entity_a": {"product": "Orbit 96", "manufacturer": "Lark Ridge", "category": "Smartwatch", "finish": "Black", "price": "1671.16"}, "entity_b": {"product": "Orbit 96", "manufacturer": "Lark Ridge", "category": "Smartwatch", "finish": "Black", "price": "1671.16"}}	Is the category the same in both records?	[ "yes", "no" ]	0	[ 0.954, 0.046 ]
956ec4bc6f3c7728	entity_matching	em	procedural	score	{"entity_a": {"name": "config_test", "maintainer": "Zutux", "registry": "pypi", "latest_version": "v6.13.0", "licence": "BSD-3-Clause"}, "entity_b": {"name": "config_test", "maintainer": "Zutux", "registry": "RubyGems", "latest_version": "6.13.0", "licence": "new bsd"}}	Are these two product records about the same real-world product?	[ "Different products.", "Related products (a variant, version or edition of each other); possibly the same.", "The same product." ]	1	[ 0.03583, 0.92834, 0.03583 ]
93a7854fcc966d1e	entity_matching	em	procedural	score	{"entity_a": {"package": "mock_crypto", "publisher": "Jeaquond", "registry": "RubyGems", "version": "v8.19.3", "license": "Apache 2.0"}, "entity_b": {"package": "mock_crypto", "publisher": "Jeaquond", "registry": "RubyGems", "version": "8.19.3", "license": "Apache License 2.0"}}	Two product listings from different sources. Do they refer to the same product?	[ "Different products.", "Related products (a variant, version or edition of each other); possibly the same.", "The same product." ]	2	[ 0.0106, 0.0424, 0.947 ]
a46cc7e1e966db54	entity_matching	em	procedural	score	{"entity_a": {"name": "Nisas-cli", "publisher": "laskork", "ecosystem": "npm", "version": "v4.26.1"}, "entity_b": {"name": "geo_date", "publisher": "laskork", "ecosystem": "PyPI", "version": "version 8.29.18", "licence": "GNU GPL v3"}}	Do `entity_a` and `entity_b` describe the same product?	[ "No match: distinct products.", "Partial match: the same line or family, but a different variant, size, version, edition or location.", "Full match: the same product, despite formatting differences." ]	0	[ 0.9246, 0.07143, 0.00397 ]
06ce409f77e5eeef	entity_matching	em	procedural	verify	{"record_a": {"blend": "HEARTH BLEND", "roastery": "Thonepa Roastery LLC", "country": "Peru", "roast_level": "Light", "process": "Dry-processed", "weight": "454g", "price": "US$12.05"}, "record_b": {"blend": "Hearth Blend", "roastery": "Thonepa Roastery", "country": "Peru", "process": "Dry-processed", "weight": "454g",...	Do `record_a` and `record_b` come from the same roaster?	[ "yes", "no" ]	0	[ 0.958, 0.042 ]
a53593945f035911	entity_matching	em	procedural	verify	{"entity_a": {"title": "Sonic G1", "brand": "Tealley", "product_type": "Notebook", "storage": "256 GB", "colour": "Sand"}, "entity_b": {"title": "grearim arc vh-73", "brand": "Grearim Corp.", "product_type": "Speaker", "colour": "Silver"}}	Are both products of the same category?	[ "yes", "no" ]	1	[ 0.08, 0.92 ]
2a790d3b81f24cf9	entity_matching	em	procedural	verify	{"entity_a": {"product": "Cedarmarsh Prism G17", "brand": "Cedarmarsh", "type": "Monitor", "color": "Sand", "price": "1920.67"}, "entity_b": {"product": "Cedarmarsh Prism G17", "brand": "Cedarmarsh", "type": "Monitor", "color": "Sand", "price": "1920.67"}}	Setting formatting aside, do the two records carry the same name?	[ "yes", "no" ]	0	[ 0.963, 0.037 ]
e7bcb909fb88cf7f	entity_matching	em	procedural	score	{"entity_a": {"book_title": "clock of seasons", "author": "Teodor Vogel", "subject": "children's books", "publisher": "roshaflas books", "format": "HC", "edition": "2e", "year": 1991}, "entity_b": {"book_title": "A Cartographer in Skosteal", "author": "AYLA YILMAZ", "subject": "Sci-fi", "publisher": "otterridge press",...	Two book listings from different sources. Do they refer to the same book?	[ "They refer to two different real-world entities.", "They refer to closely related entities that may or may not be the same one.", "They refer to one and the same real-world entity." ]	0	[ 0.948, 0.0416, 0.0104 ]
13c7fc956b223ddf	entity_matching	em	procedural	verify	{"entity_a": {"paper_title": "FEFLITH: SCALABLE CARDINALITY ESTIMATION OVER SCIENTIFIC ARCHIVES — EXTENDED ABSTRACT", "author_list": "L. Fischer, T. Rahimi", "published_in": "IEEE Trans. Knowl. Data Eng.", "year": "2013"}, "entity_b": {"paper_title": "Feflith: Scalable Cardinality Estimation over Scientific Archives", ...	Do these listings describe one and the same paper? True if: One and the same paper. False if: Two different papers, or two variants of one.	[ "yes", "no" ]	1	[ 0.088, 0.912 ]
5af7c22a8be7b89b	entity_matching	em	procedural	verify	{"entity_a": {"title": "Core Q338", "manufacturer": "Cloverwood Ltd.", "product_type": "Printers", "color": "Rose Gold", "price": "345.92 USD"}, "entity_b": {"title": "Cloverwood Core Q338", "manufacturer": "Cloverwood", "product_type": "printer", "color": "Silver", "price": "US$345.92", "listing_id": "16043"}}	Do the two entities state the same product name?	[ "yes", "no" ]	0	[ 0.941, 0.059 ]
1bac16a834999bc5	entity_matching	em	procedural	score	{"record_a": {"legal_name": "MOSS RIDGE FOODS INCORPORATED", "headquarters": "Austin, TX", "sector": "Food production", "symbol": "MIED", "url": "https://www.mossridgefoods.io/", "founded": 1972}, "record_b": {"legal_name": "Moss Ridge Energy", "headquarters": "Toronto, ON", "sector": "Utilities & Energy", "symbol": "T...	Judge whether `record_a` and `record_b` are duplicate entries for one company, ignoring formatting differences.	[ "Different companies.", "Related companies (a variant, version or edition of each other); possibly the same.", "The same company." ]	0	[ 0.92505, 0.07101, 0.00394 ]
70f261f66051d1e7	entity_matching	em	procedural	verify	{"a": {"name": "Casa Skuplark", "street": "1305 Union Avenue", "city": "Zürich", "type": "French cuisine"}, "b": {"name": "Casa Skuplark", "street": "2330 Harrow Pl.", "city": "Austin, TX", "telephone": "+1 583 508 8773", "type": "french"}}	Is the cuisine the same in both records?	[ "yes", "no" ]	0	[ 0.951, 0.049 ]
9a8f388ac2210d4f	entity_matching	em	procedural	verify	{"entity_a": {"colour": "White", "capacity": "128 GB", "category": "Tablet", "product": "Sonic M38", "price": "USD 72.28", "sku": "catalog-2", "maker": "Cloverworth Incorporated"}, "entity_b": {"product": "sonic 64", "maker": "Brepux", "category": "Headphones", "colour": "Forest Green", "price": "$ 423.67"}}	Do `entity_a.maker` and `entity_b.maker` name the same brand?	[ "yes", "no" ]	1	[ 0.073, 0.927 ]
f44ebaec16d57f9c	entity_matching	em	procedural	verify	{"entity_a": {"name": "Stoneford Foods AG", "city": "Milano", "sector": "Food production", "symbol": "NASDAQ:SNE"}, "entity_b": {"name": "Stoneford Foods AG", "city": "Milano", "sector": "Food production", "symbol": "NASDAQ:SNE"}}	Do `entity_a` and `entity_b` give the same name, ignoring case, punctuation, abbreviations and small spelling slips?	[ "yes", "no" ]	0	[ 0.965, 0.035 ]
53061d09888ed649	entity_matching	em	procedural	score	{"entity_a": {"coffee": "Night Owl Espresso", "roaster": "Ceazork Coffee", "country": "Peru", "roast_level": "Cinnamon", "processing": "Natural", "net_weight": "2.2 lb", "price": "$21.64"}, "entity_b": {"net_weight": "2.2 lb", "roast_level": "Medium", "coffee": "hearth espresso", "country": "Yemen", "processing": "Pulp...	Entity resolution: how do `entity_a` and `entity_b` relate?	[ "distinct", "related variant", "same" ]	0	[ 0.957, 0.0344, 0.0086 ]
530ccfcf7c743f78	entity_matching	em	procedural	verify	{"record_a": {"title": "flaiflyst orbit m199", "maker": "Flaiflyst", "category": "Headphones", "color": "Graphite", "price": "722.39"}, "record_b": {"title": "flaiflyst orbit m199", "maker": "Flaiflyst", "category": "headphones", "color": "Graphte"}}	Same name? Compare `record_a.title` with `record_b.title`.	[ "yes", "no" ]	0	[ 0.956, 0.044 ]
48196396de6ef279	entity_matching	em	procedural	verify	{"entity_a": {"book_title": "The Cartographer and the Archive", "authors": "Nikolai Petrov", "category": "Mystery", "publisher": "Gojiotend Books Ltd", "edition": "1st Ed.", "year": 1992}, "entity_b": {"book_title": "The Cartographer and the Archive", "authors": "Nikolai Petrov", "category": "Mysteries", "publisher": "...	Do the two entities agree on the edition?	[ "yes", "no" ]	1	[ 0.047, 0.953 ]
516f550b91c13fba	entity_matching	em	procedural	verify	{"entity_a": {"name": "Kathis Sport", "manufacturer": "Zozam", "body_style": "MPV", "year": "2021", "trim": "Premium", "engine": "3.5 L"}, "entity_b": {"name": "Kathis Sport", "manufacturer": "zozam", "body_style": "People carrier", "year": "2021", "trim": "Premium", "engine": "3.5 L"}}	Is the make the same in both records?	[ "yes", "no" ]	0	[ 0.921, 0.079 ]
1ac598619febe6d7	entity_matching	em	procedural	verify	{"left": {"name": "Grycryvol", "generic_name": "saiskeaoxetine", "strength": "25mg", "form": "Cream"}, "right": {"name": "VIPEVAINDOL", "labeler": "cloverford pharma", "generic_name": "vupongliptin", "strength": "40 MG", "form": "Cream"}}	Are `left` and `right` the same medicine, not just a related variant?	[ "yes", "no" ]	1	[ 0.076, 0.924 ]
b5e1243e29668786	entity_matching	em	procedural	verify	{"entity_a": {"product": "Raven Hill Sonic Z475", "maker": "raven hill, inc.", "product_type": "Smartwatch", "price": "820.00 USD"}, "entity_b": {"maker": "Moss Grove, Inc.", "price": "US$1722.63", "capacity": "256 GB", "product": "Sonic R62", "product_type": "Smartphones"}}	Is the brand the same in both records?	[ "yes", "no" ]	1	[ 0.056, 0.944 ]
8cb07f9d617f3c25	entity_matching	em	procedural	score	{"entity_a": {"name": "Vajiox", "company": "Ashcombe Pharma Ltd.", "generic_name": "zodavolol", "dose": "25 MG", "form": "Inhaler"}, "entity_b": {"name": "Vajiox", "company": "Ashcombe Pharma Ltd", "generic_name": "zodavolol", "dose": "25 MG", "form": "Inhaler"}}	Judge whether `entity_a` and `entity_b` are duplicate entries for one medicine, ignoring formatting differences.	[ "They describe two different medicines.", "They describe closely related medicines that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same medicine." ]	2	[ 0.0108, 0.0432, 0.946 ]
e7d9d4133fab76e4	entity_matching	em	procedural	score	{"record_a": {"book_title": "the shore and the sparrow", "authors": "Daniel Brandt", "genre": "Biography", "edition": "1st ed.", "pub_year": "2025"}, "record_b": {"book_title": "The Engine And The Orchard", "genre": "History", "imprint": "Falaix House", "edition": "1st ed.", "pub_year": "2014"}}	Entity resolution: how do `record_a` and `record_b` relate?	[ "distinct", "related variant", "same" ]	0	[ 0.92408, 0.07193, 0.004 ]
b54354a73301fa6f	entity_matching	em	procedural	score	{"entity_a": {"paper_title": "Grevai: Approximate Schema Matching over Data Lakes", "author_list": "Zara Bianchi", "venue": "ieee trans. knowl. data eng.", "year": "2017"}, "entity_b": {"paper_title": "grevai: approximate schema matching over data lakes (extended version)", "author_list": "Z. Bianchi", "venue": "VLDB"}...	How do the two entity descriptions relate as papers?	[ "No match: distinct papers.", "Partial match: the same line or family, but a different variant, size, version, edition or location.", "Full match: the same paper, despite formatting differences." ]	1	[ 0.02634, 0.94732, 0.02634 ]
c9fb52cc6a9705af	entity_matching	em	procedural	score	{"entity_a": {"name": "Ember Crossing Clos", "estate": "Preaneam Vineyards", "grape": "Tinto Fino", "vintage": "2022", "region": "Alta Serra"}, "entity_b": {"name": "Kiteton Old Vine", "estate": "Preaneam Vineyards Limited", "grape": "Tempranillo", "vintage": 2017, "alcohol": "12.50 %", "region": "Côte De Vrell", "volu...	Are these two product records about the same real-world product?	[ "distinct", "related variant", "same" ]	0	[ 0.918, 0.072, 0.01 ]
e1fa016fc7f3249a	entity_matching	em	procedural	score	{"entity_a": {"name": "Chez Rafael", "addr": "1252 Belmont Ave.", "city": "Nashville", "tel": "984/385-1217", "cuisine": "Korean"}, "entity_b": {"name": "Chez Rafael", "addr": "1252 Belmont Avenue", "city": "Nashville, TN", "tel": "+1 984 385 1217", "cuisine": "Korean"}}	Two restaurant listings from different sources. Do they refer to the same restaurant?	[ "distinct", "related variant", "same" ]	2	[ 0.0102, 0.0408, 0.949 ]
3c7094026b3ec8c7	entity_matching	em	procedural	score	{"record_a": {"company": "Quulast Pharma Holdings Ltd", "headquarters": "Vienna", "sector": "pharma", "url": "https://quulast.com", "founded": "1954"}, "record_b": {"company": "Quulast Pharma", "headquarters": "Bombay", "sector": "Biotech & Pharma", "symbol": "LSE:QUAT", "url": "https://quulast.com", "founded": "1954"}...	Two company listings from different sources. Do they refer to the same company?	[ "distinct", "related variant", "same" ]	1	[ 0.03128, 0.93744, 0.03128 ]
6b2d52b62ceb9ee7	entity_matching	em	procedural	verify	{"entity_a": {"blend": "Little Tide", "roaster": "Heron Lake Roasting Company", "country": "Rwanda", "roast_level": "Light", "process": "Dry-processed", "weight": "16 oz", "price": "US$11.69"}, "entity_b": {"blend": "Little Tide", "roaster": "Heron Lake Roasting Company", "country": "Rwandan", "roast_level": "Light roa...	Do `entity_a` and `entity_b` agree on the origin?	[ "yes", "no" ]	0	[ 0.956, 0.044 ]
4f2521d543640397	entity_matching	em	procedural	verify	{"entity_a": {"name": "greatios estate", "roaster": "Plaiwest Roasting Company Ltd", "country": "Colombia", "roast": "Light", "weight": "454g", "price": "US$15.07"}, "entity_b": {"name": "wiflados estate", "roaster": "Galemere Coffee Company", "country": "Costa Rica", "roast": "Dark roast", "process": "fully washed", "...	Do both records point to one specific product? True if: Both records describe one specific product; differences are only formatting, missing fields or price. False if: They are different products, or different variants (size, version, edition, vintage, location) of one line.	[ "yes", "no" ]	1	[ 0.06, 0.94 ]
f21ccfb5a624bdfb	entity_matching	em	procedural	verify	{"entity_a": {"product": "Granite Crossing Orbit 20", "manufacturer": "Granite Crossing Ltd", "type": "smart watch", "price": "USD 980.05"}, "entity_b": {"product": "Greares Orbit E275", "type": "smart watch", "finish": "Forest Green", "price": "US$435.00"}}	Same name? Compare `entity_a.product` with `entity_b.product`.	[ "yes", "no" ]	1	[ 0.052, 0.948 ]
d67b4d0e1a363b52	entity_matching	em	procedural	score	{"entity_a": {"model": "Stubru", "brand": "Jeskior Automotive", "body_style": "Hatch", "year": "2020", "grade": "EX", "engine": "2000 cc", "listing_id": "48517"}, "entity_b": {"model": "Stubru", "brand": "Jeskior Automotive", "body_style": "5-door hatchback", "year": "2020", "engine": "3000 cc"}}	Judge whether `entity_a` and `entity_b` are duplicate entries for one product, ignoring formatting differences.	[ "They describe two different products.", "They describe closely related products that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same product." ]	1	[ 0.03625, 0.92751, 0.03625 ]
3d0a7ef32e010823	entity_matching	em	procedural	score	{"entity_a": {"name": "Trestal Reserve Lot", "roastery": "Larch Peak Coffee Co.", "origin": "Kenyan", "roast_level": "light roast", "processing": "fully washed", "net_weight": "1.1 lb", "price": "US$40.30"}, "entity_b": {"price": "$ 40.30", "origin": "Kenya", "processing": "Fully washed", "net_weight": "500g", "name": ...	Are these two coffee records about the same real-world coffee?	[ "Different coffees.", "Related coffees (a variant, version or edition of each other); possibly the same.", "The same coffee." ]	1	[ 0.02593, 0.94814, 0.02593 ]
d5547cb8d95aaef7	entity_matching	em	procedural	verify	{"entity_a": {"name": "Casa Sheha", "street": "2185 Calloway Drive", "city": "København", "telephone": "515-598-8394", "cuisine": "Mexican"}, "entity_b": {"city": "Krakow", "cuisine": "American cuisine", "name": "Adrian's Eatery", "street": "542 Mercer Rd."}}	Does `entity_b` state the same city as `entity_a`?	[ "yes", "no" ]	1	[ 0.043, 0.957 ]
c981d77854d6fe16	entity_matching	em	procedural	score	{"entity_a": {"title": "Dance With Me", "artist": "Felix Nakamura", "album_name": "Broken Feathers", "genre": "Heavy metal", "length": "05:49", "price": "0.99 USD"}, "entity_b": {"title": "Dance With Me (Remix)", "artist": "Felix Nakamura", "album_name": "Broken Feathers", "genre": "Heavy metal", "length": "6m 22s", "...	How do the two entity descriptions relate as tracks?	[ "They describe two different tracks.", "They describe closely related tracks that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same track." ]	1	[ 0.03335, 0.93331, 0.03335 ]
5c854104477c4940	entity_matching	em	procedural	verify	{"left": {"package": "quamith-token", "maintainer": "Stifleth", "registry": "RubyGems", "version": "6.15.4", "licence": "GPLv3"}, "right": {"package": "jojupand-path", "maintainer": "mudrym", "registry": "npm", "version": "version 5.28.19", "licence": "GPL-3.0"}}	Do `left` and `right` agree on the registry?	[ "yes", "no" ]	1	[ 0.064, 0.936 ]
6da831d6a02ca27a	entity_matching	em	procedural	score	{"entity_a": {"restaurant": "Tobias's Oyster Bar", "street": "167 Larkin Rd", "city": "Prague", "phone": "229.518.4065", "type": "Greek / Mediterranean"}, "entity_b": {"restaurant": "Tobias's Oyster Bar", "street": "2595 Kingsley Ln.", "city": "Austin, TX", "phone": "2293992087", "type": "greek / mediterranean"}}	How do the two entity descriptions relate as restaurants?	[ "Different restaurants.", "Related restaurants (a variant, version or edition of each other); possibly the same.", "The same restaurant." ]	1	[ 0.03874, 0.92253, 0.03874 ]
c33d6f51cf38a63a	entity_matching	em	procedural	verify	{"entity_a": {"wine": "Domaine Lyglest Rouge", "estate": "Jasum", "grape": "Merlot", "vintage": 2013, "alcohol": "13.0 % vol", "region": "Coastal Ranges", "volume": "375 ml"}, "entity_b": {"wine": "Domaine Lyglest Rouge", "estate": "Jasum", "grape": "Merlot", "vintage": 2013, "alcohol": "13.0% ABV"}}	Do `entity_a` and `entity_b` give the same name, ignoring case, punctuation, abbreviations and small spelling slips?	[ "yes", "no" ]	0	[ 0.955, 0.045 ]
440d008ef70d7074	entity_matching	em	procedural	score	{"entity_a": {"book_title": "the season's archive", "author": "Henrik Farouk", "category": "Fantasy fiction", "imprint": "Briarmere Books", "binding": "Digital edition", "edition": "3rd ed."}, "entity_b": {"book_title": "The Season's Archive", "author": "Farouk, Henrik", "category": "Fantasy", "imprint": "Briarmere Boo...	Two product listings from different sources. Do they refer to the same product?	[ "They describe two different products.", "They describe closely related products that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same product." ]	1	[ 0.03335, 0.93331, 0.03335 ]
d0fe5a3d16428407	entity_matching	em	procedural	verify	{"entity_a": {"title": "Wild Clocks", "writer": "Ingrid Vogel", "subject": "Nonfiction / History", "imprint": "TANSY PEAK HOUSE", "edition": "1st edition", "published": "2007"}, "entity_b": {"title": "wild clocs", "imprint": "Tansy Peak House", "format": "HC", "edition": "1st edition", "published": "2007"}}	Do `entity_a.edition` and `entity_b.edition` state the same edition, allowing for units and formatting?	[ "yes", "no" ]	0	[ 0.925, 0.075 ]
919489bf5894633e	entity_matching	em	procedural	verify	{"entity_a": {"title": "Vapynt N67", "maker": "Tansy Cove Ltd", "category": "headphone", "finish": "Sand", "price": "USD 1924.87"}, "entity_b": {"title": "Core E255", "maker": "Tansy Cove Limited", "category": "Headphone", "finish": "rose gold"}}	Setting formatting aside, do the two records carry the same name?	[ "yes", "no" ]	1	[ 0.066, 0.934 ]
fe7dda28d2b22bea	entity_matching	em	procedural	score	{"entity_a": {"title": "Approximate Indexing for Key-Value Stores", "author_list": "Quintero, A.", "venue": "Annual Meeting of the Association for Computational Linguistics"}, "entity_b": {"title": "Approximate Indexing for Key-Value Stores: Extended Version", "author_list": "Quintero, A.", "venue": "ACM Transactions o...	Are these two paper records about the same real-world paper?	[ "They refer to two different real-world entities.", "They refer to closely related entities that may or may not be the same one.", "They refer to one and the same real-world entity." ]	1	[ 0.02963, 0.94073, 0.02963 ]
bc3814cbc9119897	entity_matching	em	procedural	score	{"entity_a": {"title": "gibry 82", "manufacturer": "Librufla Corporation", "category": "Smartphones", "storage": "256gb", "colour": "forest green", "price": "$ 1040.08"}, "entity_b": {"title": "Librufla Flux F91", "manufacturer": "Librufla", "category": "Keyboard", "colour": "Black", "price": "$ 1358.11"}}	Compare the two records. How closely do they match as products?	[ "Different products.", "Related products (a variant, version or edition of each other); possibly the same.", "The same product." ]	0	[ 0.9252, 0.07087, 0.00394 ]
6da343d6fad6b609	entity_matching	em	procedural	score	{"entity_a": {"company": "Prakoth Provisions GmbH", "city": "Copenhagen", "sector": "Food & Beverage", "ticker": "PRO", "website": "https://prakoth.de", "founded": "1994"}, "entity_b": {"company": "Prakoth Provisions Holdings Inc.", "city": "Osaka", "sector": "Food & Beverage", "website": "https://www.prakoth.de/", "fo...	Entity resolution: how do `entity_a` and `entity_b` relate?	[ "They refer to two different real-world entities.", "They refer to closely related entities that may or may not be the same one.", "They refer to one and the same real-world entity." ]	1	[ 0.03583, 0.92834, 0.03583 ]
78973881ab5eef84	entity_matching	em	procedural	verify	{"entity_a": {"blend": "Morning Roast", "roastery": "Silvermere Coffee Company Ltd.", "origin": "yemen", "process": "Fully washed", "price": "US$35.76"}, "entity_b": {"blend": "Morning Roast", "roastery": "Silvermere Coffee Company Ltd.", "origin": "yemen", "process": "Fully washed", "price": "US$35.76"}}	Is the product named the same in both records?	[ "yes", "no" ]	0	[ 0.959, 0.041 ]
4c955c6ce523782f	entity_matching	em	procedural	score	{"entity_a": {"name": "Lumeen 15", "brand": "Skeajast", "product_type": "External solid-state drive", "capacity": "512 GB", "colour": "White", "price": "$1409.46"}, "entity_b": {"name": "Skeajast Lumen 15", "brand": "skeajast", "product_type": "External SSD", "capacity": "512gb", "colour": "White", "price": "1409.46"}}	How likely is it that both records describe one and the same product?	[ "Different products.", "Related products (a variant, version or edition of each other); possibly the same.", "The same product." ]	2	[ 0.0136, 0.0544, 0.932 ]
b90b12e6319b9966	entity_matching	em	procedural	verify	{"entity_a": {"brand_name": "Trepistol", "ingredient": "sudaindlukast", "dose": "0.1 g", "dosage_form": "tabs", "source": "feed_b"}, "entity_b": {"brand_name": "Trepistol", "manufacturer": "Clovercombe Pharma", "ingredient": "sudaindlukast", "dose": "10 MG", "dosage_form": "tabs"}}	Do these listings describe one and the same medicine?	[ "yes", "no" ]	1	[ 0.079, 0.921 ]
56476acb7c609e5c	entity_matching	em	procedural	verify	{"entity_a": {"title": "Dunecombe Bavuceas AC-35", "maker": "Dunecombe Corp.", "type": "Notebook", "capacity": "2 TB", "color": "Midnight Blue", "price": "USD 902.95"}, "entity_b": {"title": "Bavuceas AC-34", "maker": "dunecombe corporation", "capacity": "2 TB", "color": "Midnight Blue", "sku": "scrape"}}	Setting formatting aside, do the two records carry the same name?	[ "yes", "no" ]	1	[ 0.072, 0.928 ]
00a5e15cd9105c62	entity_matching	em	procedural	verify	{"entity_a": {"name": "Bellwell Telecom Incorporated", "headquarters": "St. Louis", "sector": "Telecommunications", "ticker": "SIX: BLL", "domain": "https://www.bellwell.co/", "founded": "1931"}, "entity_b": {"name": "Bellwell Telecom Incorporated", "headquarters": "St. Louis", "sector": "Telecommunications", "ticker":...	Do the two entities describe the same industry?	[ "yes", "no" ]	0	[ 0.96, 0.04 ]
312bfe64470e1c67	entity_matching	em	procedural	score	{"entity_a": {"name": "Cedararsh Provisions", "city": "Madrid", "sector": "Food & Beverage", "symbol": "NASDAQ: CEH", "url": "cedarmarsh.net", "founded_year": "1923"}, "entity_b": {"name": "Cedarmarsh Provisions PLC", "city": "Madrid", "sector": "Food & Beverage", "symbol": "CEH", "url": "https://www.cedarmarsh.net/", ...	Are these two company records about the same real-world company?	[ "They describe two different companies.", "They describe closely related companies that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same company." ]	2	[ 0.0088, 0.0352, 0.956 ]
e22643db1d0e30b8	entity_matching	em	procedural	verify	{"entity_a": {"name": "Beatriz Weiss", "affiliation": "tansy ridge institute", "role": "POSTDOCTORAL FELLOW", "discipline": "History", "city": "Mumbai"}, "entity_b": {"name": "B. Weiss", "affiliation": "Tansy Ridge Institute", "role": "Postdcotoral Fellow", "discipline": "history", "city": "Mumbai"}}	Do `entity_a` and `entity_b` give the same name, ignoring case, punctuation, abbreviations and small spelling slips?	[ "yes", "no" ]	0	[ 0.938, 0.062 ]
ddbdda2a831f2a63	entity_matching	em	procedural	score	{"entity_a": {"name": "Gale Field Outfitters", "city": "Wien", "industry": "retail", "ticker": "TSX: GIE", "url": "https://galefield.de"}, "entity_b": {"name": "Gale Field Circuits", "city": "København", "industry": "Semiconductors", "ticker": "NASDAQ:GFD", "url": "http://galefieldcircuits.com"}}	How do the two entity descriptions relate as companies?	[ "Different companies.", "Related companies (a variant, version or edition of each other); possibly the same.", "The same company." ]	0	[ 0.92415, 0.07186, 0.00399 ]
6030e4221b595d91	entity_matching	em	procedural	verify	{"entity_a": {"package": "colorx", "maintainer": "omarerik", "registry": "nuget", "latest_version": "6.15.10"}, "entity_b": {"package": "colorx", "maintainer": "omarerik", "registry": "RubyGems", "latest_version": "6.15.10 (stable)", "license": "GNU GPL v3"}}	Setting formatting aside, do the two records carry the same name?	[ "yes", "no" ]	0	[ 0.949, 0.051 ]
e74fa1cf10d37689	entity_matching	em	procedural	score	{"record_a": {"title": "Lumen K358", "brand": "Floquomu", "product_type": "Keyboards", "color": "Black", "price": "$ 1388.90"}, "record_b": {"title": "Lumen K358", "brand": "Floquomu Ltd.", "product_type": "Computer keyboard", "price": "US$1388.90"}}	How likely is it that both records describe one and the same product?	[ "They refer to two different real-world entities.", "They refer to closely related entities that may or may not be the same one.", "They refer to one and the same real-world entity." ]	2	[ 0.011, 0.044, 0.945 ]
c8208921bbdd4306	entity_matching	em	procedural	verify	{"entity_a": {"name": "mock_image", "maintainer": "Falcon Lake", "registry": "npmjs", "latest_version": "v8.14.8", "licence": "Mozilla Public License 2.0"}, "entity_b": {"name": "mock_image", "maintainer": "Falcon Lake", "registry": "npmjs", "latest_version": "v8.14.8"}}	Same product?	[ "yes", "no" ]	0	[ 0.944, 0.056 ]
2c80700738dc5dc8	entity_matching	em	procedural	score	{"record_a": {"full_name": "Stefan Kariuki", "organization": "Branark Institute", "role": "Lecturer", "discipline": "History", "city": "St. Paul"}, "record_b": {"full_name": "Kariuki, S.", "organization": "university of rejoprant", "role": "Senior Engineer", "discipline": "Chemistry", "city": "NYC"}}	Compare the two records. How closely do they match as people?	[ "No match: distinct people.", "Partial match: the same line or family, but a different variant, size, version, edition or location.", "Full match: the same person, despite formatting differences." ]	1	[ 0.03459, 0.93082, 0.03459 ]
daf20c7f3b8f1d90	entity_matching	em	procedural	verify	{"entity_a": {"founded_year": "1891", "url": "zecroth.com", "company": "Zecroth Aviation UK", "hq": "Zurich", "sector": "aerospace"}, "entity_b": {"company": "Zecroth Aviation Limited", "hq": "Milan", "sector": "Aerospace", "symbol": "ZET", "url": "http://zecroth.com", "founded_year": "1891"}}	Do these listings describe one and the same company?	[ "yes", "no" ]	1	[ 0.089, 0.911 ]
14bdd7851dca1627	entity_matching	em	procedural	score	{"a": {"restaurant": "Hequioth Jonas", "addr": "892 Fairmount Drive", "city": "München", "telephone": "229-904-4759", "category": "Korean BBQ"}, "b": {"restaurant": "Hequioth Oyster Bar", "addr": "1203 Wexford Dr.", "city": "Praha", "telephone": "+1 279 363 8131", "category": "Indian cuisine"}}	Judge whether `a` and `b` are duplicate entries for one restaurant, ignoring formatting differences.	[ "They describe two different restaurants.", "They describe closely related restaurants that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same restaurant." ]	0	[ 0.939, 0.0488, 0.0122 ]
65704cd5d786232d	entity_matching	em	procedural	score	{"a": {"restaurant": "aisha's bistro", "address": "477 Harrow Drive", "city": "Melbourne", "type": "vietnamese"}, "b": {"restaurant": "CHIARA'S BISTRO", "address": "1746 Prescott Lane", "city": "Portland, Maine", "phone": "886.955.0297", "type": "Thai"}}	Compare the two records. How closely do they match as restaurants?	[ "They describe two different restaurants.", "They describe closely related restaurants that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same restaurant." ]	0	[ 0.933, 0.0536, 0.0134 ]
dca3ba43d2a12f92	entity_matching	em	procedural	verify	{"record_a": {"name": "Wild Letters: A Novel", "author": "O. Lindqvist", "genre": "Business", "publisher": "Faflaind House", "binding": "Trade paperback", "published": 2003}, "record_b": {"name": "Wild Voyages", "author": "Engel, Bruno", "genre": "Cookery", "publisher": "Oak Field Books", "binding": "paperback", "editi...	Is the genre the same in both records?	[ "yes", "no" ]	1	[ 0.064, 0.936 ]
5a4b7e0654131e9d	entity_matching	em	procedural	score	{"entity_a": {"title": "Ragly: Incremental Retrieval over Time Series", "authors": "Wanjiru Engel and René Jansen and Felix Novak", "venue": "KDD"}, "entity_b": {"title": "Ragly: Incremental Retrieval over Time Series — Extended Abstract", "authors": "Engel, W.; Jansen, R.; Novak, F.", "venue": "ACM Trans. Database Sys...	Judge whether `entity_a` and `entity_b` are duplicate entries for one paper, ignoring formatting differences.	[ "No match: distinct papers.", "Partial match: the same line or family, but a different variant, size, version, edition or location.", "Full match: the same paper, despite formatting differences." ]	1	[ 0.02511, 0.94978, 0.02511 ]
993e9fa9c04b2a6e	entity_matching	em	procedural	score	{"entity_a": {"package": "traceify", "owner": "jonasnils", "registry": "nuget.org", "latest_version": "3.26.9"}, "entity_b": {"package": "Traceify", "owner": "Wrenmoor", "registry": "Maven Central", "latest_version": "0.27.6"}}	Do `entity_a` and `entity_b` describe the same product?	[ "Different products.", "Related products (a variant, version or edition of each other); possibly the same.", "The same product." ]	0	[ 0.92585, 0.07024, 0.0039 ]
b3056424d15e8b70	entity_matching	em	procedural	score	{"entity_a": {"blend": "HEARTH ESPRESSO", "roastery": "Clover Ridge Roastery L.L.C.", "origin": "brazil", "roast_level": "Dark roast", "processing": "Pulped natural", "net_weight": "340g", "price": "17.47 USD"}, "entity_b": {"blend": "Hearth Espresso", "roastery": "Clover Ridge Roastery L.L.C.", "origin": "Brazilian", ...	Judge whether `entity_a` and `entity_b` are duplicate entries for one coffee, ignoring formatting differences.	[ "Different coffees.", "Related coffees (a variant, version or edition of each other); possibly the same.", "The same coffee." ]	2	[ 0.0102, 0.0408, 0.949 ]
826aa7a1902dce46	entity_matching	em	procedural	score	{"record_a": {"product": "Lumen M576", "maker": "Flofelost", "type": "Cell phone", "colour": "Rose Gold", "price": "356.03"}, "record_b": {"colour": "Graphite", "product": "Lumen P27", "maker": "Emberfell", "type": "Computer monitor", "price": "US$1009.07"}}	Entity resolution: how do `record_a` and `record_b` relate?	[ "They describe two different products.", "They describe closely related products that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same product." ]	0	[ 0.953, 0.0376, 0.0094 ]
f5528d08f7ebf41b	entity_matching	em	procedural	verify	{"entity_a": {"model": "dushyrk cross", "manufacturer": "Gleaplajax Motors", "body": "Coupé", "year": "2010", "trim": "base", "engine": "1.0 L"}, "entity_b": {"model": "Moprem Cross", "manufacturer": "pudrork", "body": "Minivan", "year": "2017", "trim": "Limited", "engine": "1.5 litre"}}	Does the same make stand behind both records?	[ "yes", "no" ]	1	[ 0.073, 0.927 ]
e678b318d4ff1002	entity_matching	em	procedural	verify	{"entity_a": {"name": "K70", "brand": "Hopukyr", "type": "Minivan", "year": "2005", "grade": "LX", "displacement": "1.5 litre"}, "entity_b": {"name": "l30", "brand": "hopukyr", "type": "Crossover SUV", "year": 2019, "grade": "S", "displacement": "1.2 litre"}}	Do the two records agree on the make?	[ "yes", "no" ]	0	[ 0.932, 0.068 ]
66ecab6702a5d605	entity_matching	em	procedural	verify	{"a": {"name": "Winter Feathers", "authors": "T. Okafor", "genre": "Poetry", "imprint": "Bramble Field House", "binding": "Ebook", "year": "2004"}, "b": {"name": "Meadow of Meadows", "genre": "Science Fiction", "binding": "Paperback", "edition": "1e", "year": 2011}}	Would merging the two records be correct?	[ "yes", "no" ]	1	[ 0.057, 0.943 ]
54de736ea3e4bab6	entity_matching	em	procedural	score	{"entity_a": {"title": "Hidden Meadows", "author": "Nordin, Erik", "genre": "Children's books", "publisher": "Ottermoor Books", "edition": "1st edition", "published": "1998"}, "entity_b": {"title": "The Forest's Clock", "author": "Reyes, N.", "genre": "Poetry", "edition": "3rd Ed.", "published": 1986}}	How do the two entity descriptions relate as products?	[ "They refer to two different real-world entities.", "They refer to closely related entities that may or may not be the same one.", "They refer to one and the same real-world entity." ]	0	[ 0.945, 0.044, 0.011 ]
84a597136a6f1b89	entity_matching	em	procedural	verify	{"a": {"name": "GRYWO SINGLE", "roastery": "Tansy Point Roasting Company", "origin": "Rwandan", "roast": "French roast", "processing": "Fully washed", "size": "454g", "price": "43.60 USD"}, "b": {"name": "Labu Single Origin", "roastery": "Maple Ridge Roasting Company", "origin": "Peruvian", "roast": "City", "processing...	Ignoring formatting and units, is the roast level identical in both records?	[ "yes", "no" ]	1	[ 0.075, 0.925 ]
15afced324e7a71c	entity_matching	em	procedural	score	{"left": {"product": "Fleceafork 8", "manufacturer": "Hollowton", "product_type": "Wi-Fi router", "finish": "Graphite", "price": "1215.94 USD"}, "right": {"product": "Fleceaforrk 8", "manufacturer": "hollowton", "product_type": "Wireless router", "finish": "Graphite", "price": "US$1215.94"}}	Are these two product records about the same real-world product?	[ "They describe two different products.", "They describe closely related products that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same product." ]	2	[ 0.0116, 0.0464, 0.942 ]
9ea6860d4a3c16d5	entity_matching	em	procedural	verify	{"entity_a": {"song": "Clock Waltz ( Live )", "artist": "Leaskanest", "album_name": "Crimson Forests", "genre": "Country", "length": "4:45", "price": "USD 1.29"}, "entity_b": {"song": "clock waltz", "artist": "Leaskanest", "genre": "Country", "length": "04:27", "price": "1.29"}}	Ignoring formatting and units, is the duration identical in both records?	[ "yes", "no" ]	1	[ 0.073, 0.927 ]
98156c862bf5cc26	entity_matching	em	procedural	verify	{"entity_a": {"name": "Nipream", "make": "Rariont Automotive", "type": "hatch", "year": 2018, "grade": "Base"}, "entity_b": {"name": "Staplaith Sport", "make": "Measkastai", "type": "Pickup", "year": "2020", "engine": "1.0L"}}	Do the two entities state the same car name?	[ "yes", "no" ]	1	[ 0.069, 0.931 ]
af796876e5c87701	entity_matching	em	procedural	score	{"entity_a": {"product": "pebblestone nova 62", "manufacturer": "Pebblestone Ltd", "product_type": "Camera", "colour": "Graphite", "price": "$587.25"}, "entity_b": {"product": "pebblestone nova 62", "manufacturer": "Pebblestone Ltd", "product_type": "Camera", "colour": "Graphite", "price": "$587.25"}}	How do the two entity descriptions relate as products?	[ "They describe two different products.", "They describe closely related products that may or may not be the same one: a variant, a special edition, or a name that could plausibly refer to either.", "They describe one and the same product." ]	2	[ 0.0082, 0.0328, 0.959 ]
b2fe8e03c3fa9fc6	entity_matching	em	procedural	verify	{"entity_a": {"title": "On the Limits of Entity Resolution in Data Lakes", "author_list": "Imani Wojcik, Ayla Hoang, Viktor Rossi and Teodor Weiss", "venue": "TKDE"}, "entity_b": {"author_list": "Imani Wojcik et al.", "venue": "kdd", "title": "Distributde Join Processing For Relational Databases", "year": 2022}}	Do `entity_a` and `entity_b` give the same name, ignoring case, punctuation, abbreviations and small spelling slips?	[ "yes", "no" ]	1	[ 0.066, 0.934 ]
ee70c2eb732b245c	entity_matching	em	procedural	verify	{"left": {"restaurant": "The Painted Tide", "addr": "1342 Harrow Rd", "city": "Denver, CO", "phone": "404/934-0687", "type": "Greek / Mediterranean"}, "right": {"restaurant": "Mirror & Tide", "addr": "496 Kingsley St", "city": "Portland, OR", "phone": "815-474-3663", "type": "Indian"}}	Do `left.type` and `right.type` name the same cuisine?	[ "yes", "no" ]	1	[ 0.072, 0.928 ]
d28ba656157c9c5b	entity_matching	em	procedural	score	{"left": {"label": "Broseam Meadow", "producer": "Juniperwick Estate", "grape": "Tempranillo", "vintage": "2011", "alcohol": "13.0% ABV", "appellation": "coastal ranges", "bottle": "1.5 litre", "source": "scrape"}, "right": {"label": "Broseam Meadow", "producer": "juniperwick estate", "grape": "Tempranillo", "vintage"...	How do the two entity descriptions relate as wines?	[ "Different wines.", "Related wines (a variant, version or edition of each other); possibly the same.", "The same wine." ]	2	[ 0.0104, 0.0416, 0.948 ]
96a2679d32246c32	entity_matching	em	procedural	verify	{"entity_a": {"region": "Kettle Valley", "volume": "0.75 L", "abv": "ABV 12.0%", "year": "2012", "wine": "Château Jabashar", "winery": "Driovov", "grape": "Primitivo"}, "entity_b": {"wine": "château jabashar", "winery": "Driovov", "grape": "Zinfandel", "abv": "ABV 12.0%", "region": "Kettle Valley"}}	Does the same winery stand behind both records?	[ "yes", "no" ]	0	[ 0.942, 0.058 ]
9ac51e49aff66b16	entity_matching	em	procedural	verify	{"entity_a": {"name": "silent voyage", "roaster": "Otterwell Coffee Roasters", "country": "Indonesia (Sumatra)", "roast_level": "French roast", "processing": "Sun-dried natural", "size": "32 oz", "price": "US$42.17", "listing_id": "feed_b"}, "entity_b": {"name": "Silent Voyage", "roaster": "Otterwell Coffee Roasters", ...	Do `entity_a.name` and `entity_b.name` hold the same name?	[ "yes", "no" ]	0	[ 0.957, 0.043 ]
f8869c0279711a6d	entity_matching	em	procedural	score	{"entity_a": {"title": "Bonyv K469", "maker": "Waglur Corp", "product_type": "Smart Watch", "colour": "Silver", "price": "$ 1629.50"}, "entity_b": {"title": "Bonyv K469", "maker": "Waglur Corp", "product_type": "Smart Watch", "colour": "Silver", "price": "$ 1629.50"}}	How do the two entity descriptions relate as products?	[ "distinct", "related variant", "same" ]	2	[ 0.0096, 0.0384, 0.952 ]
98a10aec66906b64	entity_matching	em	procedural	verify	{"left": {"name": "Oscar Brandt", "affiliation": "rush ridge stores inc", "title": "Chief Economist", "discipline": "Legal", "city": "Lyon, France"}, "right": {"name": "Bruno Brandt", "affiliation": "Elmwell Institute", "title": "Senior Engineer", "discipline": "Physics", "city": "Portland, Maine"}}	Are the affiliation values equivalent?	[ "yes", "no" ]	1	[ 0.055, 0.945 ]
e1c0d654bf4f9393	entity_matching	em	procedural	score	{"entity_a": {"paper_title": "On the Limits of Cardinality Estimation in Product Catalogs", "authors": "Farouk, L.", "venue": "ACL", "year": "2000"}, "entity_b": {"paper_title": "towards learned entity resolution", "authors": "Lukas Farouk", "venue": "sigmod conference", "year": "2019"}}	Entity resolution: how do `entity_a` and `entity_b` relate?	[ "They refer to two different real-world entities.", "They refer to closely related entities that may or may not be the same one.", "They refer to one and the same real-world entity." ]	0	[ 0.918, 0.072, 0.01 ]
0da3a57e9d341fed	entity_matching	em	procedural	score	{"a": {"company": "Hollow Grove Power Ltd.", "hq": "Nashville, TN", "sector": "utilities & energy", "symbol": "HLV", "url": "https://www.hollowgrove.net/", "founded_year": 2003}, "b": {"company": "Hollow Grove Power Holdings", "hq": "Copenhagen", "sector": "Utilities & Energy", "url": "www.hollowgrove.net", "founded_ye...	Judge whether `a` and `b` are duplicate entries for one company, ignoring formatting differences.	[ "Different companies.", "Related companies (a variant, version or edition of each other); possibly the same.", "The same company." ]	1	[ 0.04082, 0.91837, 0.04082 ]
340ea44377376455	entity_matching	em	procedural	score	{"entity_a": {"name": "logio", "maintainer": "arjunmbek", "ecosystem": "npmjs", "latest_version": "3.11.18 (stable)", "licence": "Mozilla Public License 2.0"}, "entity_b": {"name": "diff_crypto", "maintainer": "Gruvom", "ecosystem": "rubygems.org", "latest_version": "version 4.22.17", "licence": "GPLv3"}}	How likely is it that both records describe one and the same product?	[ "No match: distinct products.", "Partial match: the same line or family, but a different variant, size, version, edition or location.", "Full match: the same product, despite formatting differences." ]	0	[ 0.921, 0.0632, 0.0158 ]
3ebca12a721e2c6b	entity_matching	em	procedural	verify	{"entity_a": {"name": "graph_plot", "owner": "idrislind", "ecosystem": "rubygems.org", "version": "version 1.10.16", "license": "apache 2.0"}, "entity_b": {"name": "litimage", "owner": "jaifleflel", "ecosystem": "npm", "version": "0.12.14 (stable)", "license": "MIT"}}	Do `entity_a` and `entity_b` agree on the registry?	[ "yes", "no" ]	1	[ 0.067, 0.933 ]
51fe60aba3a092b0	entity_matching	em	procedural	verify	{"entity_a": {"blend": "hearth roast", "roastery": "Grustedry Roasting Company L.L.C.", "origin": "Rwandan", "roast": "Cinnamon", "process": "Wet-processed", "size": "907 g", "price": "$18.41"}, "entity_b": {"blend": "Paflashork Reserve Lot", "roastery": "Styquiofi Roastery", "origin": "Yemeni", "roast": "City", "proce...	Would a cataloguer read the two names as one and the same name?	[ "yes", "no" ]	1	[ 0.053, 0.947 ]
4f32f3cb0ea9f846	entity_matching	em	procedural	score	{"entity_a": {"label": "Tefex Garden", "producer": "Trioprirk Winery", "variety": "Chenin Blanc", "vintage": "2020", "alcohol": "15.0% ABV", "region": "Rive Gauche", "bottle": "37.5 cl"}, "entity_b": {"label": "Tefex Gardne", "producer": "Trioprirk Winery Ltd", "variety": "Steen", "vintage": "2020", "alcohol": "15.0%",...	Entity resolution: how do `entity_a` and `entity_b` relate?	[ "distinct", "related variant", "same" ]	2	[ 0.011, 0.044, 0.945 ]
55eb201785fdfd2c	entity_matching	em	procedural	verify	{"entity_a": {"coffee": "Negun Cregioth Origin", "brand": "Crane Field Roastery", "origin": "Guatemalan", "roast_level": "city", "process": "natural", "net_weight": "2 lbs", "price": "36.10 USD"}, "entity_b": {"coffee": "cregioth single origin", "brand": "Drefyr Coffee Limited", "origin": "sumatra", "roast_level": "cin...	Do `entity_a.brand` and `entity_b.brand` name the same roaster?	[ "yes", "no" ]	1	[ 0.071, 0.929 ]
1cf3be8e1648294b	entity_matching	em	procedural	verify	{"entity_a": {"product": "Triovi 89", "manufacturer": "Rookdale", "type": "Portable speaker", "finish": "Sand", "price": "$2041.22"}, "entity_b": {"product": "Triovi 89", "manufacturer": "Rookdale Corp", "type": "Speaker", "finish": "Sand", "price": "USD 2041.22"}}	Are the two entities from the same brand?	[ "yes", "no" ]	0	[ 0.939, 0.061 ]
a4c9c7c32607a884	entity_matching	em	procedural	score	{"entity_a": {"name": "miogrant t39", "manufacturer": "pebble point", "category": "Tablet", "finish": "sand", "price": "$ 979.02"}, "entity_b": {"name": "PEBBLE POINT NOQUAX T47", "category": "smartphones", "storage": "1 tb", "price": "875.55 USD"}}	Judge whether `entity_a` and `entity_b` are duplicate entries for one product, ignoring formatting differences.	[ "distinct", "related variant", "same" ]	0	[ 0.92423, 0.07178, 0.00399 ]
9c6f325aab220c16	entity_matching	em	procedural	verify	{"a": {"song": "kitchen anthem", "artist": "Wanond", "album_name": "Glass Bridges", "genre": "EDM", "time": "3m 47s", "price": "$1.29"}, "b": {"song": "Kitchen Slow", "source": "89604", "genre": "Rap", "price": "$0.99", "artist": "THE PAINTED MAPS"}}	Do the records give one genre, possibly written differently?	[ "yes", "no" ]	1	[ 0.079, 0.921 ]
d9870fc122e8f35d	entity_matching	em	procedural	verify	{"entity_a": {"coffee": "Burninng River", "roastery": "Juniperhurst Coffre Co. LLC", "country": "Ethiopian", "roast": "City", "processing": "Natural", "price": "$14.04"}, "entity_b": {"coffee": "burning river", "roastery": "Juniperhurst Coffee Company Llc", "country": "ethiopian", "roast": "City", "processing": "dry-pr...	Same roast level?	[ "yes", "no" ]	0	[ 0.92, 0.08 ]

End of preview. Expand in Data Studio

Decima Knowledge-Graph Judgments

About 200k judgments a knowledge-graph pipeline makes after extraction, with exact or two-model-agreed labels: are these two records the same entity? Is this extracted fact asserted, hypothetical, denied or only planned? Does the document discuss this relation at all, and in which direction? What does this mention refer to? Are these two things alternatives?

Today these checks are LLM calls, one per edge or candidate pair. This dataset trains a small local model to make them: decima-agent 2.2 was trained on it and, as a drop-in for Jev in William Lyon's public knowledge-graph notebooks, catches 11 of 11 planted hedged facts (Jev: 8) and matches the Beer entity pairs at AUC 0.972 (Jev: 0.992). None of the notebooks' documents or pairs are in this dataset.

Built by A. M. Madani (@amyrmahdy). Generators: teacher/kg_proc.py and teacher/kg_llm.py.

Configs

config	train	test	state → decision
`entity_matching`	50,000	1,638	two records (products, companies, people, papers, places…) → same / related variant / different, plus per-field checks
`mention_pairs`	24,859	814	two mentions with their sentences → same entity? abbreviation or ticker? (with and without an ontology rule)
`modality`	25,038	806	a sentence → asserted / hypothetical / negated / forward-looking; can it be recorded as a fact?
`relation_pairs`	33,392	742	two things → alternatives / complementary / same / unrelated
`documents`	59,896	3,581	a 150–450-word business article → a fact's status, "does it discuss this?", which relation in which direction (or none), a mention's type
`support_threads`	3,825	160	a 10–16-turn support thread → intent, priority, resolved

Columns: id, config, task, origin (procedural or llm+blind-check), kind (choose, verify, score), state, question, choices, gold, probs (the soft target the model trained on). Questions are written the way TypeSafe choice / noul / score requests render, including the wording of TypeSafe's entity-alignment cookbook.

from datasets import load_dataset
ds = load_dataset("amyrmahdy/decima-kg-judgments", "modality", split="test")

How the labels were made

procedural: exact, from code. Records and mentions are generated with controlled differences (case, punctuation, abbreviations, legal suffixes, typos, missing fields, tickers, versions, sizes, vintages); modality sentences from templates with controlled cue words; relation pairs from hand-written role lists, including hard negatives such as two tools with different roles.
llm+blind-check. Code invents a small fictional world and the facts with their modality; Gemma 4 (26B-A4B, run locally) writes the article or thread around them; a separate call that sees only the text and the question labels it blind, and a row is kept only when that answer matches the code's label (status 96 %, type 97 % agreement).

All companies, people and products are invented. Test splits hold out record domains, industries, cue words, question phrasings, article styles, support channels and one ontology, so they measure transfer, not memory.

Limitations

Synthetic text: cleaner and more formulaic than real news, filings or support logs.
llm+blind-check labels are only as good as two calls of the same model agreeing.
Some labelling choices are conventions (a product's variant is "related", not "same"; a ticker without an ontology rule is the issuer). They are documented in the generators.

License

CC BY 4.0. Gemma 4 is Apache-2.0.

Downloads last month: 106

Collection including amyrmahdy/decima-kg-judgments

Decima: small local judges

Collection

Small open CPU models: multilingual PII detector, LLM guard, agent gates, decision models. Arabic, Persian, English. • 16 items • Updated 1 day ago