Datasets:

yoh
/

modular-sentence-encoders-sts

lang_a stringclasses 4 values	lang_b stringclasses 4 values	sentence_a stringlengths 8 174	sentence_b stringlengths 8 163	score float64 1 5
en	en	There is no boy playing outdoors and there is no man smiling	A group of kids is playing in a yard and an old man is standing in the background	3.3
en	en	A group of boys in a yard is playing and a man is standing in the background	The young boys are playing outdoors and the man is smiling nearby	3.7
en	en	A group of children is playing in the house and there is no man standing in the background	The young boys are playing outdoors and the man is smiling nearby	3
en	en	A brown dog is attacking another animal in front of the tall man in pants	A brown dog is attacking another animal in front of the man in pants	4.9
en	en	A brown dog is attacking another animal in front of the man in pants	A brown dog is helping another animal in front of the man in pants	3.665
en	en	Two dogs are wrestling and hugging	There is no dog wrestling and hugging	3.3
en	en	A brown dog is attacking another animal in front of the man in pants	There is no dog wrestling and hugging	2.7
en	en	Two dogs are wrestling and hugging	A brown dog is attacking another animal in front of the tall man in pants	2.9
en	en	Two dogs are wrestling and hugging	A brown dog is helping another animal in front of the man in pants	2.3
en	en	A person in a black jacket is doing tricks on a motorbike	A man in a black jacket is doing tricks on a motorbike	4.9
en	en	There is no man in a black jacket doing tricks on a motorbike	A person in a black jacket is doing tricks on a motorbike	3.6
en	en	A person in a black jacket is doing tricks on a motorbike	A person on a black motorbike is doing tricks with a jacket	3
en	en	A skilled person is riding a bicycle on one wheel	A person is riding the bicycle on one wheel	4.3
en	en	Nobody is riding the bicycle on one wheel	A person is riding the bicycle on one wheel	4.1
en	en	A person is riding the bicycle on one wheel	There is no man in a black jacket doing tricks on a motorbike	3.2
en	en	A person is riding the bicycle on one wheel	A person in a black jacket is doing tricks on a motorbike	3.3
en	en	A man with a jersey is dunking the ball at a basketball game	There is no man dunking the ball at a basketball game	4
en	en	The game of basketball consists of a ball being dunked by a man with a jersey	A man with a jersey is dunking the ball at a basketball game	4.2
en	en	A man who is playing dunks the basketball into the net and a crowd is in background	The player is dunking the basketball into the net and a crowd is in background	4.7
en	en	The player is missing the basket and a crowd is in background	The player is dunking the basketball into the net and a crowd is in background	3.9
en	en	A man with a jersey is dunking the ball at a basketball game	The player is missing the basket and a crowd is in background	3.5
en	en	The player is dunking the basketball into the net and a crowd is in background	The ball is being dunked by a man with a jersey at a basketball game	4
en	en	There is no man dunking the ball at a basketball game	The player is dunking the basketball into the net and a crowd is in background	3.5
en	en	The game of basketball consists of a ball being dunked by a man with a jersey	The player is dunking the basketball into the net and a crowd is in background	3.6
en	en	Two people are kickboxing and spectators are watching	Two people are fighting and spectators are watching	4.4
en	en	Two spectators are kickboxing and some people are watching	Two people are kickboxing and spectators are watching	4
en	en	Two people are kickboxing and spectators are watching	Two young women are sparring in a kickboxing fight	3.9
en	en	Two women are sparring in a kickboxing match	Two people are fighting and spectators are watching	3.7
en	en	Two spectators are kickboxing and some people are watching	Two women are sparring in a kickboxing match	3.6
en	en	Two people are kickboxing and spectators are watching	Two women are sparring in a kickboxing match	3.9
en	en	Kids in red shirts are playing in the leaves	Children in red shirts are playing in the leaves	4.9
en	en	Children in red shirts are playing in the leaves	Children in red shirts are sleeping in the leaves	4
en	en	Children in red shirts are playing in the leaves	Children covered by leaves are playing with red shirts	3.9
en	en	Three boys are jumping in the leaves	Children in red shirts are playing in the leaves	4.1
en	en	Kids in red shirts are playing in the leaves	Three kids are jumping in the leaves	3.9
en	en	Children in red shirts are sleeping in the leaves	Three kids are jumping in the leaves	3.1
en	en	Children covered by leaves are playing with red shirts	Three kids are jumping in the leaves	3.9
en	en	Two children are lying in the snow and are drawing angels	Two children are lying in the snow and are making snow angels	4.4
en	en	Two children are lying in the snow and are making snow angels	There is no child lying in the snow and making snow angels	3.8
en	en	Two people wearing snowsuits are on the ground making snow angels	Two people in snowsuits are lying in the snow and making snow angels	4.8
en	en	Two people in snowsuits are lying in the snow and making snow angels	Nobody in snowsuits is lying in the snow and making snow angels	4.1
en	en	Two children are lying in the snow and are making snow angels	Two people wearing snowsuits are on the ground making snow angels	4.6
en	en	Two children are lying in the snow and are making snow angels	Nobody in snowsuits is lying in the snow and making snow angels	3.6
en	en	There is no child lying in the snow and making snow angels	Two people in snowsuits are lying in the snow and making snow angels	3.3
en	en	A little girl is looking at a woman in costume	A young girl is looking at a woman in costume	4.7
en	en	A little girl is looking at a woman in costume	The little girl is looking at a man in costume	3.8
en	en	A little girl is looking at a woman in costume	A little girl in costume looks like a woman	2.9
en	en	Masked people are looking in the same direction in a forest	A little girl is looking at a woman in costume	1.3
en	en	A little girl is looking at a woman in costume	People wearing costumes are gathering in a forest and are looking in the same direction	2.6
en	en	A man is jumping into an empty swimming pool	A man is jumping into an empty pool	4.8
en	en	A lone biker is jumping in the air	A man is jumping into an empty swimming pool	1.6
en	en	A man is jumping into an empty pool	A biker is jumping in the air, alone	1.6
en	en	A man is jumping into an empty pool	There is no biker jumping in the air	1.2
en	en	Four girls are doing backbends and playing outdoors	Four girls are doing backbends and playing indoors	4
en	en	Four girls are doing backbends and playing in the garden	Four children are doing backbends in the park	3.7
en	en	Four girls are doing backbends and playing indoors	Four children are doing backbends in the park	3.7
en	en	Four kids are doing backbends in the park	Four girls are doing backbends and playing outdoors	4.3
en	en	Four girls are doing backbends and playing outdoors	Four children are doing backbends in the park	4.1
en	en	A player is throwing the ball	A player is running with the ball	3.5
en	en	Two groups of people are playing football	Two teams are competing in a football match	4.7
en	en	Two teams are competing in a football match	Two teams are competing in a baseball game	3.9
en	en	Two teams are competing in a football match	Football is being played by two competing teams	4.8
en	en	A player is running with the ball	Football is being played by two competing teams	3
en	en	Two teams are competing in a football match	A man who is playing is running with the ball in his hands	2.7
en	en	Five children are standing in front of a wooden hut	Five children are standing in front of a hut made of wood	4.8
en	en	Five kids are standing close together and one kid has a gun	Five kids are standing close together and one kid has a weapon	4.8
en	en	Five children are standing in front of a wooden hut	Five kids are standing close together and one kid has a weapon	2.8
en	en	Five children are standing in front of a hut made of wood	Five kids are standing close together and one kid has a gun	3
en	en	Five children are standing in front of a wooden hut	Five kids are standing close together and one kid has a gun	2.3
en	en	A person is wearing a hat and is sitting on the grass	There is no person wearing a hat and sitting on the grass	3.3
en	en	A person is sitting in a field and is wearing a hat	A man is sitting in a field	4
en	en	A man is sitting in a field	There is no person wearing a hat and sitting on the grass	2.7
en	en	An old man is sitting in a field	A person is wearing a hat and is sitting on the grass	2.8
en	en	A person is wearing a hat and is sitting on the grass	A man is running in a field	1.4
en	en	This group of people is practicing water safety and wearing preservers	Preservers are being worn by a group of people who are practicing water safety	4.9
en	en	Nobody is practicing water safety and wearing preservers	This group of people is practicing water safety and wearing preservers	3.4
en	en	A group of friends are riding the current in a raft	Preservers are being worn by a group of people who are practicing water safety	2.9
en	en	Nobody is practicing water safety and wearing preservers	A group of friends are riding the current in a raft	2.2
en	en	A group is not riding the current in a raft	This group of people is practicing water safety and wearing preservers	2.8
en	en	A deer is jumping over a fence	A deer is jumping over the enclosure	4.535
en	en	A wild deer is jumping a fence	A deer is jumping a fence	4.8
en	en	A deer is jumping a fence	There is no deer jumping fences	4.1
en	en	A wild deer is jumping a fence	A deer is jumping over a fence	4.5
en	en	There is no deer jumping fences	A deer is jumping over a fence	2.685
en	en	A deer is jumping over the enclosure	A deer is jumping a fence	4.2
en	en	A deer is jumping a fence	A deer isn't jumping over the fence	3.55
en	en	A deer is jumping over a fence	A deer is jumping a fence	4.9
en	en	People are walking outside a building that has many murals on it	People are walking outside the building that has several murals on it	4.7
en	en	Several people are in front of a building which is covered by colors	Several people are in front of a colorful building	4.7
en	en	Several people are in front of a building which is covered by colors	People are walking outside a building that has many murals on it	3.8
en	en	Several people are in front of a colorful building	People are walking inside a building that has many murals on it	3.4
en	en	People are walking outside a building that has many murals on it	Several people are in front of a colorful building	3.7
en	en	A boy is hitting a baseball	A child is hitting a baseball	4.4
en	en	A child is hitting a baseball	A child is missing a baseball	3.8
en	en	A family is watching a little boy who is hitting a baseball	A family is watching a boy who is hitting a baseball	4.6
en	en	A little boy is watching a family that is hitting a baseball	A family is watching a little boy who is hitting a baseball	3.7
en	en	A family is watching a boy who is hitting a baseball	A child is hitting a baseball	3.7
en	en	A little boy is watching a family that is hitting a baseball	A child is hitting a baseball	3.7
en	en	A family is watching a little boy who is hitting a baseball	A boy is hitting a baseball	4.2
en	en	A family is watching a little boy who is hitting a baseball	A child is hitting a baseball	4.1

End of preview. Expand in Data Studio

STS / STR evaluation sets, resolved (23 languages)

The semantic textual similarity and relatedness evaluation data of Modular Sentence Encoders: Separating Language Specialization from Cross-Lingual Alignment (ACL 2025), in one flat schema.

This repository contains no new data. It is a redistribution of six existing benchmarks, with the alignment work already applied: pairing monolingual sets into cross-lingual ones by row index, intersecting SICK pair IDs across translations, matching KorSTS rows to the STS benchmark, combining the STS17 extension columns, and transliterating the Uyghur data from Cyrillic to Arabic script. Doing that at load time is where evaluation code usually goes quietly wrong, so it is frozen into the data instead.

Usage

from datasets import load_dataset

ds = load_dataset("yoh/modular-sentence-encoders-sts", "sts17", split="test")
de_fr = ds.filter(lambda r: r["lang_a"] == "de" and r["lang_b"] == "fr")

Schema

One row per sentence pair, in every config:

column	type	meaning
`lang_a`	string	language of `sentence_a`, a two-letter code
`lang_b`	string	language of `sentence_b`
`sentence_a`	string	first sentence
`sentence_b`	string	second sentence
`score`	float	gold similarity or relatedness; the scale is the source benchmark's

Each config has a single test split: these are evaluation sets, and none of them is meant to be trained on.

A monolingual set is simply a row where lang_a == lang_b. Score scales differ between configs (0–5 for STS and SICK, 0–1 for STR), which does not matter for the Spearman correlation these sets are used with, but does matter if you pool them.

config	language pairs	rows	scale
`sts17`	30	10,096	0–5
`stsb`	48	38,388	0–5
`sick`	16	78,496	1–5
`str24`	6 (monolingual only)	4,191	0–1

Two sizes depart from the round numbers of the source benchmarks, both inherited from the original resolution: STS17 ko-ko is the full KorSTS test set (2,846 rows, not 250), and the twelve STSB pairs involving Korean have 799 rows rather than 800, because one row could not be matched.

Sources

config	languages	built from
`sts17`	ar, cs, de, en, es, fr, it, ko, nl, tr	SemEval-2017 Task 1 (Cer et al., 2017) and its extended multilingual version (Reimers & Gurevych, 2020); Korean from KorSTS
`stsb`	az, en, kk, ky, ko, ug, uz	the STS benchmark; Turkic languages from Kardeş-NLU (Senel et al., 2024); Korean from KorSTS (Ham et al., 2020)
`sick`	en, es, nl, pl	SICK (Marelli et al., 2014), SICK-NL (Wijnholds & Moortgat, 2021), SICK-PL (Dadas et al., 2020), SICK-ES (Huertas-García et al., 2022)
`str24`	am, en, ha, mr, rw, te	SemRel2024, SemEval-2024 Task 1 (Ousidhoum et al., 2024)

Cross-lingual pairs in sts17, stsb and sick are formed by taking sentence_a from one language's copy of a row and sentence_b from another's, which is valid because those benchmarks are translations of the same items. The gold score is the source benchmark's score for that row.

Licensing

The configs do not share one license, because their sources do not:

config	terms
`sick`	SICK is CC BY-NC-SA 3.0; its translations are derivatives and inherit it. SICK-NL's code is MIT, its data is not.
`sts17`, `stsb`	The similarity scores of the STS benchmark family are CC BY-SA 4.0; the underlying sentences keep the licenses of the corpora they were drawn from. KorSTS is CC BY-SA 4.0. Kardeş-NLU states no license.
`str24`	SemRel2024 states no license in its repository or on the Hub.

The repository is labelled CC BY-NC-SA 3.0, the strictest of the terms that are actually stated, and is intended for non-commercial research use. Where a source states no license, redistribution here is on the understanding that the data was published for open evaluation use; if you are one of those authors and prefer that we link rather than redistribute, please open a discussion and the config will be replaced by a resolution script.

Please cite the source benchmarks you evaluate on, not only this repository.

Citation

@inproceedings{huang-etal-2025-modular,
    title = "Modular Sentence Encoders: Separating Language Specialization from Cross-Lingual Alignment",
    author = "Huang, Yongxin and Wang, Kexin and Glava{\v{s}}, Goran and Gurevych, Iryna",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    year = "2025",
    pages = "2167--2187",
    url = "https://aclanthology.org/2025.acl-long.108/",
}

Code: https://github.com/UKPLab/acl2025-modular-sentence-encoders

Downloads last month: 50