Hugging Face's logo

Datasets:
mteb
/
sts15-sts

Tasks:
Sentence Similarity
Modalities:
Text
Formats:
json
Sub-tasks:
semantic-similarity-scoring
Languages:
English
Size:
1K - 10K
ArXiv:
Tags:
mteb
text
Libraries:
Datasets
pandas
License:
Dataset card Data Studio Files
xet
Community
1
Dataset Viewer
Auto-converted to Parquet
split
stringclasses
1 value
sentence1
stringlengths
15
302
sentence2
stringlengths
16
275
score
float64
0
5
test
you 'll need to check the particular policies of each publisher to see what is allowed and what is not allowed .
if you need to publish the book and you have found one publisher that allows it .
3
test
i am not aware of any university run participant pools .
at the universities i 've worked in north america , human-subject studies ( esp .
0.6
test
apart from admission ( which normally should not be related to age ) , there is something that 's related to your question .
age is not a phd admission facor in the continental europe , too .
1
test
this post refers to research in the stem fields , and may not be applicable to other research topics .
this is about my experience in computer engineering i found that reading for breadth was the more important approach .
0.6
test
if you don 't want to pay for adobe acrobat pro , as @ schultem mentions , latex can do this with embedded hyperlinks .
if you don 't mind hosting your files online , slideshare is a good solution .
0.8
test
i don 't think there are likely to be any standards that address this issue specifically .
you 're going to find answers all over the map for this one ( i.e. , there probably aren 't " standards " ) .
2.4
test
my answer would be depending on which gre are you referring to ?
the problem i see with the gres is that the scoring range is highly compressed .
1
test
if you have access to it , you can very easily do that with thomson reuters ’ web of science portal .
you can have a look at the arxiv cultoromics website at http : / / arxiv.culturomics.org , which however searches in articles on the arxiv .
0.2
test
in general there are a number of different distinct workflows that lead to a poster .
for completeness , apple 's pages has quite a few nice poster layouts .
0.6
test
it is perfectly okay to ask for the figure knowing well that the other person may not give .
you can always ask , then it is the choice of the author to accept or not .
3.6
test
the academic year does start around september in the usa and i think most european countries .
i would not accelerate things , to avoid getting worse grades that you want .
0.2
test
i will advise exactly the contrary of what bravo just said in another answer : go for a !
you could defer admission , but it 's a little unusual to defer for a year .
0.2
test
edit , again : the question seems to have changed , yet again .
i can think of three reasons why a faculty member might reveal how a student performed on a qualifying examination .
0
test
i agree with seteropere completely , " network science " is a very broad subject .
i would say you are approaching it in the wrong way .
0
test
supporting an ideology in an ostentatious manner ( rather than privately or discreetly ) is a personality problem .
i think for an appropriate answer i 'd need to know exactly what the ' bad ' ideology is .
1.2
test
this does not fully answer your question , but it is certainly one consideration .
i am a phd student in computational science , which is an interdisciplinary major spanning mathematics , computer science , and engineerning .
0
test
a few possible problems : 1 ) such an id system seems a little intrusive .
in mathematics , we already have a database of this sort - two them , in fact , with in mathscinet and zentralblatt math .
0.4
test
as an complement to the other answers , let me address the question of why planets tend towards tidal locking .
the other answers here are fantastic at explaining in a technical sense .
0.4
test
stars are rotating , but that 's not the cause for their stability .
stars rotate due to the angular momentum of the gas they formed from .
3
test
it indeed appears the andromeda galaxy ( m31 ) and the milky way ( mw ) are en route to a collision .
in a few billion years , the milky way and andromeda will collide .
4.4
test
the short answer is : we are made from the " stuff " contributed by many stars .
it is highly unlikely that we 're made out of stuff from only one star .
3.2
test
all we know is this : distant objects are receding from us at a rate proportional to their distance from us .
the expansion of space means that objects in cosmological distances are receding away from each other .
3.4
test
like chris said above , stars rotate to conserve their angular momentum .
stars are rotating , but that 's not the cause for their stability .
3.4
test
according to this website the peak visible magnitude will be about 10.5 around february 2nd .
the aavso data seems to indicate that it might have already peaked , at around 10.5 ( visual ) .
3.6
test
all of these effects are related to the 11 year solar cycle .
i will be specific to answering the part regarding the prediction of solar flares .
0.8
test
we are moving " ... relative to the comoving cosmic rest frame ... at some 371 km / s towards the constellation leo " .
there is no ' still ' that is not relative to some other object .
0.8
test
yes , it 's true that the apparent size of the moon is 30 arcmin .
when you gaze at the moon " live " , you are not seeing a still image .
1
test
let us assume the data for a globular cluster to be equivalent to that of m13 .
globular clusters occupy an interesting place in the spectrum of composite stellar systems .
1.6
test
why is the speed of light 299,792,458 m / s , and not ( for instance ) 3,1 or 4,3 x 10 ^ 44 m / s ?
speed of light being finite is one of the fundamentals of our universe .
1
test
as per your definition , a star formed in a binary system ( hence near some gravitating body ) is a generalised planet .
the main flaw with your definition of a " generalised planet " would be the boundary between stars and planets .
3.2
test
a method used to calculate the distance between stars is 3 dimensional trigonometry .
you only need two-dimensional trigonometry if you know the distances to the two stars and their angular separation .
2
test
the farther away , the faster the galaxies move away from us .
here is my answer to a similar question posted on the physics stack exchange website .
0
test
stars form in star-formation regions , which itself develop from molecular clouds .
it may be possible for a solar system like ours to exist outside of a galaxy .
0.6
test
a black hole ( briefly ) , is nothing more than a dead star whose mass was more than 3 solar masses .
i think you mean to ask , " what is the local sidereal time of greenwhich ? " .
0
test
yes , but it does not stay in the same constellation .
constellations are human constructs to make sense of the night sky .
1.4
test
asteroids are what are called minor planets - small inner solar system rocky bodies .
the objects you are refering to are actually two different objects : asteroids and comets .
2
test
here is a lexical decision task i wrote years ago using php and javascript ( github link ) .
just for reference , i ended up writing simple web-based lexical decision task software myself .
2.6
test
it 's pretty difficult to imagine a person with social anxiety disorder being an extrovert .
on the surface , it does seem like social anxiety disorder and extroversion shouldn 't both exist in the same person .
4.2
test
it is hard to delineate emotion completely from thoughts or sensations , because emotion can contain both .
technically speaking , an emotion is not different from thought the way it is processed in brain , both involve neurotransmitters .
2.8
test
i 've found an additional concept that may be related to the original question : prototype theory .
according to current models of human concept learning , the answer to your question is both .
0
test
there are individuals who possess extraordinary memory ability , sometimes called eidetic memory .
there are many arguments for why this is not the case .
0
test
spike trigger is a specific type or you could say a sub-set of reverse correlations , covariance and probabilities .
there 's the naïve version of spike triggered averaging , and the sophisticated version .
2
test
as mentioned in the other comments , anova is problematic when mixing types of predictor variables .
i like to think of multitasking as rapid task switching .
0
test
q : what is the reason for people to implicitly trust their peers in extreme ( or not ) situations ?
one human-being reliance on other ( s ) awareness is defined by the level of trust between the humans .
2.6
test
intrinsic reward is the emotional pay-off that you get by completing an activity .
according to frederick herzberg 's two-factor theory of motivation , intrinsic rewards are motivators ( e.g .
2.6
test
presumably the decision of drivers to slow down in response to work zone signage is influenced by many factors .
this short talk deals with issues of " cheating slightly " : dan ariely : our buggy moral code .
0
test
when studying adult iq , general adult norms are often used .
basically , what you do is you compare each age group iq to the standardization sample .
2.6
test
some ancient historical precedent exists for preferring $ 10 $ , but also for $ 6 $ , so that 's mixed support from wikipedia on perfect numbers .
this is indeed possible , but i haven 't seen it done experimentally for reasons other than feasibility .
0
test
these are interesting studies , but the bottom line is : yes , environment affects cognitive abilities .
i 'm not sure what you mean by cognitive capacity , but i absolutely believe that language shapes the way we think .
1.6
test
since this is a relatively new problem for behavioral researchers , i don 't know that there is a common consensus .
i 'll address just the first of your three sub-questions , the others have been answered by chuck sherrington .
0
test
if non-human animals do have intelligence too , why is their intelligence not as advanced as humans ?
your own theory is bad because it seems to follow from a lamarkian interpretation of evolutionary theory .
0.6
test
the hare psychopathy checklist is often used to assess psychopathy in clinical settings .
from an article entitled can you call a 9-year-old a psychopath ?
1.8
test
i wrap steaks individually in freezer paper , then put them into ziploc freezer bags .
wrap well in plastic ( individually ) and then put in a freezer bag .
3.8
test
in addition to oil , make sure the grill is nice and hot before you start cooking .
i 'd been using paper towels to oil the grill , and i just discovered that newspapers work a lot better .
1.6
test
the best way i have found is to soak the ears in husk for several hours before grilling .
soak whole ears in the husk in a bowl of water .
3.2
test
there are actually quite a few differences between the different types of flour .
the difference is the amount of protein contained in the flour which can range from 5 % to 15 % .
2.6
test
i 've never adjusted the length of time based on number of eggs .
i don 't boil eggs enough to keep a feel for the timing ...
3
test
i just hold the cheese closer to the edge away from the grater .
by the wording of the question , i 'm pretty sure you 're using one of those pyramidal graters or a " sheet " grater .
1.4
test
i 've stored pancakes in the freezer for over two month and not had any freezer burn .
its going to depend on the type of packaging and also the freezer .
1.6
test
break the bulb into cloves , breaking off the bottom and the dry stick that goes up the middle .
lay down a knife horizontaly on the cloves and hit it .
1.6
test
i managed to find this recipe , i think this is where i got the idea .
i 've only made a couple of bbq sauces , but they 've always come out really well .
1
test
i believe one cannot taste the difference between minced ( canned ) and fresh garlic once it is cooked .
in my experience , fresh garlic is great for when you want texture , or to slow down the flavour infusion .
1.8
test
i grill salmon that i get with the skin on .
salmon is a great choice and it 's readily available pretty much everywhere .
1.8
test
the recipe i work from has you put the meat in the freezer , then pan sear it .
starting with a whole chilled piece of meat , i quickly pour boiling water over all the outside surface .
1.4
test
they don 't taste identical at all , and even more important , they don 't smell the same .
black pepper and white pepper are differing preparations of the fruit of the piper nigrum plant .
2.6
test
this depends on the shelf life of the cultivar of tomato you purchased .
tomatoes will last longer if kept in the fridge , but i actually recommend against keeping them there .
2
test
coagulated egg proteins in combination with gluten give baked goods the supporting structure they need .
there are basically two types of cakes : butter and sponge ( or génoise ) .
1
test
a common one is haddock and is in season pretty much year round .
white fish are , as a rule , more delicately flavored than darker fleshed fish .
1.4
test
good cheddar has little chunks of calcium lactate on / in it - could it be that ?
i have two suspicions : one , you 're undercooking your roux , the flour and butter mixture , and not fully incorporating the flour .
0.6
test
copper can be useful for certain tasks due to the its metallic properties ( heats quickly , distributes heat evenly , etc .
probably subjective , but ... good clad stainless steel will give you much better cooking performance per dollar than copper .
1.4
test
she obviously has a legit excuse , although she could have been more responsible and " have " told you in advance .
the first have covers both been more responsible and told you in that sentence .
0.4
test
in some cases the implication could even be that tom does not love her .
there is no implication about tom whatsoever in such a sentence .
1.8
test
i agree with others ( and yourself ) in that i have never heard this phrase in english .
i have never heard or seen the phrase you mention .
4
test
the assertion is false , at least as you 've restated it here .
" i had mainly been " is fine , in the context you 're using it .
0.6
test
i 'd suggest you look to the constructions of the native language ( s ) of your students .
as bikeboy389 said , you can learn a lot by looking at students ' native languages .
2.8
test
i see some references to status as uncountable as well , but that doesn 't make much sense to me .
there are some cases " status " may be accounted of countable .
2.6
test
if i understand you correctly , you are looking for a hypernym of inclusion and exclusion .
inclusion is the act of including and exclusion is the act of excluding .
2.6
test
firstly , you should be aware that there are two different kinds of dashes : the en dash and the em dash .
semicolon is used to join sentences that can stand alone , but are joined to emphasize their relationship .
0.6
test
pardon the brevity of this answer , but i would say " named " is preferred within the context of your example .
named is preferred in your example , since you are formally giving a name to your method .
4.4
test
in standard english , i think " don 't got " has the same register as " ain 't " .
i 'd say the expression i don 't got ... is generally rare in spoken ame .
2.2
test
the " more posh " words are usually latin ( occasionally greek ) in origin .
as shinto pointed out , vulgar words exist in every language .
1
test
i 'm american , and i 've never heard the bathroom referred to as st. john 's .
in situations like these , i tend to lengthen out the phrase .
0
test
the room used for defecation is almost always referred to by euphemism .
i 'm english , and would probably use ' toilet ' most of the time , and always in the context of a private home .
1.6
test
i can only think of context using insofar as or insofar that .
the form insofar is american english ; the standard form in british and other english is ' in so far ' .
1.8
test
as others have said , " ten of six " , though not used in many english speaking areas , would be understood as 5 : 50 .
" ten of six " probably means 5 : 50 but i have not heard it before .
3.8
test
certainly many usage guides have advised against use of this " singular they " on various " logical " grounds .
second one ( for writing purpose ) , if i believe the " they / their ( singular ) " article .
3
test
this may be overly obvious , but in american english , saying " you 're welcome " is certainly polite and standard .
i 'm not sure how " not at all " sounds in response to " thank you " .
1.4
test
back in the day , " cable " was used to describe communications sent abroad .
the word " cable " has its origins in the days of the telegraph .
3
test
those expressions could have slightly different meanings especially when talking about finance .
i suspect the phrase originated in the habit of drawing the internet as a cloud shape on network diagrams .
0.4
test
in the united states , jails are operated by cities and counties ( or equivalent ) .
jail is a municipal level , prison is on a state / federal / provincial / territory level .
3.6
test
yes , your sentence will be clearer if you write the actual word instead of the slash .
most style guides recommend against using the slash in phrases like “ x and / or y ” , as well as “ this is hard / impossible to do ” .
3.4
test
if you understand yourself and your enemy , you will be far more likely to win any battle .
another meaning behind this quote aside from probabilities of winning a theoretical battle is the importance of knowing / understanding others .
2.8
test
there are a lot of push up variations you can do and they all stress different muscle systems .
there are several different pushup variations out there and most of them provide a unique advantage .
4.2
test
you definitely do not want to be supporting your weight with your arms on the bike for normal riding .
no , don 't support your weight on your arms your hands simply aren 't really made for supporting all that weight .
4.2
test
yes , you should stretch before running , but not the sit down and reach kind of stretching .
neuromuscular activation and dynamic stretching here is an excellent video on a dynamic stretching routine for running .
1.8
test
i feel your pain - i absolutely pour with sweat with any vigorous exercise , even in cool conditions .
sweat rate is sweat rate , you can 't really control it .
1
test
running downhill is the best way to strengthen your quad muscles .
additionally often people will think a downhill marathon is easier , but in fact , is often harder than an uphill marathon .
2
test
i recently stumbled on a great article that talks about protein consumption for the purposes of building muscle ( recovery included ) .
low reps with high weights don 't add mass because the duration of the exercise is kept short .
1
test
rotate through square , lunges and calf raises doing each until you feel a little burn .
air squats , forward and side lunges , sprawl , wall chair , and lots and lots of jump rope .
1.6
test
if you 're doing a full body workout 6 days a week , you absolutely need the rest week .
there are both physiological and neurological reasons to build in rest .
2.6
End of preview. Expand in Data Studio

STS15

An MTEB dataset
Massive Text Embedding Benchmark

SemEval STS 2015 dataset

Task category t2t
Domains Blog, News, Web, Written, Spoken
Reference https://www.aclweb.org/anthology/S15-2010

How to evaluate on this task

You can evaluate an embedding model on this dataset using the following code:

import mteb

task = mteb.get_tasks(["STS15"])
evaluator = mteb.MTEB(task)

model = mteb.get_model(YOUR_MODEL)
evaluator.run(model)

To learn more about how to run models on mteb task check out the GitHub repitory.

Citation

If you use this dataset, please cite the dataset as well as mteb, as this dataset likely includes additional processing as a part of the MMTEB Contribution.


@inproceedings{bicici-2015-rtm,
  address = {Denver, Colorado},
  author = {Bi{\c{c}}ici, Ergun},
  booktitle = {Proceedings of the 9th International Workshop on Semantic Evaluation ({S}em{E}val 2015)},
  doi = {10.18653/v1/S15-2010},
  editor = {Nakov, Preslav  and
Zesch, Torsten  and
Cer, Daniel  and
Jurgens, David},
  month = jun,
  pages = {56--63},
  publisher = {Association for Computational Linguistics},
  title = {{RTM}-{DCU}: Predicting Semantic Similarity with Referential Translation Machines},
  url = {https://aclanthology.org/S15-2010},
  year = {2015},
}


@article{enevoldsen2025mmtebmassivemultilingualtext,
  title={MMTEB: Massive Multilingual Text Embedding Benchmark},
  author={Kenneth Enevoldsen and Isaac Chung and Imene Kerboua and Márton Kardos and Ashwin Mathur and David Stap and Jay Gala and Wissam Siblini and Dominik Krzemiński and Genta Indra Winata and Saba Sturua and Saiteja Utpala and Mathieu Ciancone and Marion Schaeffer and Gabriel Sequeira and Diganta Misra and Shreeya Dhakal and Jonathan Rystrøm and Roman Solomatin and Ömer Çağatan and Akash Kundu and Martin Bernstorff and Shitao Xiao and Akshita Sukhlecha and Bhavish Pahwa and Rafał Poświata and Kranthi Kiran GV and Shawon Ashraf and Daniel Auras and Björn Plüster and Jan Philipp Harries and Loïc Magne and Isabelle Mohr and Mariya Hendriksen and Dawei Zhu and Hippolyte Gisserot-Boukhlef and Tom Aarsen and Jan Kostkan and Konrad Wojtasik and Taemin Lee and Marek Šuppa and Crystina Zhang and Roberta Rocca and Mohammed Hamdy and Andrianos Michail and John Yang and Manuel Faysse and Aleksei Vatolin and Nandan Thakur and Manan Dey and Dipam Vasani and Pranjal Chitale and Simone Tedeschi and Nguyen Tai and Artem Snegirev and Michael Günther and Mengzhou Xia and Weijia Shi and Xing Han Lù and Jordan Clive and Gayatri Krishnakumar and Anna Maksimova and Silvan Wehrli and Maria Tikhonova and Henil Panchal and Aleksandr Abramov and Malte Ostendorff and Zheng Liu and Simon Clematide and Lester James Miranda and Alena Fenogenova and Guangyu Song and Ruqiya Bin Safi and Wen-Ding Li and Alessia Borghini and Federico Cassano and Hongjin Su and Jimmy Lin and Howard Yen and Lasse Hansen and Sara Hooker and Chenghao Xiao and Vaibhav Adlakha and Orion Weller and Siva Reddy and Niklas Muennighoff},
  publisher = {arXiv},
  journal={arXiv preprint arXiv:2502.13595},
  year={2025},
  url={https://arxiv.org/abs/2502.13595},
  doi = {10.48550/arXiv.2502.13595},
}

@article{muennighoff2022mteb,
  author = {Muennighoff, Niklas and Tazi, Nouamane and Magne, Lo{\"\i}c and Reimers, Nils},
  title = {MTEB: Massive Text Embedding Benchmark},
  publisher = {arXiv},
  journal={arXiv preprint arXiv:2210.07316},
  year = {2022}
  url = {https://arxiv.org/abs/2210.07316},
  doi = {10.48550/ARXIV.2210.07316},
}

Dataset Statistics

Dataset Statistics

The following code contains the descriptive statistics from the task. These can also be obtained using:

import mteb

task = mteb.get_task("STS15")

desc_stats = task.metadata.descriptive_stats

{
    "test": {
        "num_samples": 3000,
        "number_of_characters": 346442,
        "unique_pairs": 3000,
        "min_sentence1_length": 15,
        "average_sentence1_len": 57.822,
        "max_sentence1_length": 302,
        "unique_sentence1": 2942,
        "min_sentence2_length": 16,
        "average_sentence2_len": 57.65866666666667,
        "max_sentence2_length": 275,
        "unique_sentence2": 2310,
        "min_score": 0.0,
        "avg_score": 2.40591333333333,
        "max_score": 5.0
    }
}

This dataset card was automatically generated using MTEB

Downloads last month
3,182
Size of downloaded dataset files:
116 kB
Size of the auto-converted Parquet files:
212 kB
Number of rows:
3,000