Towards Best Practices for Open Datasets for LLM Training
Paper
•
2501.08365
•
Published
•
40
BigLAM: BigScience Libraries, Archives and Museums
non-profit
AI & ML interests
🤗 Hugging Face x 🌸 BigScience initiative to create open source community resources for LAMs.
Recent Activity
View all activity
biglam's activity
storytracer
authored
a
paper
2 days ago
Update README.md
#2 opened 3 days ago
by
librarian-bot
librarian-bot
updated
a
dataset
3 days ago
Update README.md
#2 opened 3 days ago
by
librarian-bot
Convert dataset to Parquet
#3 opened 3 days ago
by
davanstrien
davanstrien
updated
a
dataset
3 days ago
Convert dataset to Parquet
#1 opened 3 days ago
by
davanstrien
davanstrien
updated
a
dataset
3 days ago
Convert dataset to Parquet
#1 opened 3 days ago
by
davanstrien
davanstrien
posted
an
update
5 days ago
Post
2902
Introducing scandi-fine-web-cleaner
davanstrien/scandi-fine-web-cleaner, the first model trained on FineWeb-C community annotations!
FineWeb2 is a massive multilingual dataset for pre-training language models. Like any web-scale dataset, it contains low-quality content. How can we improve it?
Over the past months, an amazing community of 400+ annotators has been labelling content quality (using Argilla) across 23 languages through the FineWeb-C initiative.
Today, I'm happy to share the first classifier trained on this data.
🔍 What we've built:
- A lightweight classifier that efficiently removes low-quality content
- 90%+ precision demonstrated on Danish & Swedish
- Can process the 43M+ documents in Danish FineWeb2 with minimal compute
🌍 Why this matters: The approach can be reproduced for any of the 23 languages in FineWeb-C ( data-is-better-together/fineweb-c). We can improve training data quality at scale without massive compute resources by starting with community annotations and training small, efficient classifiers.
Want to build a classifier for your language? Check out the full blog post with code examples and implementation details: https://danielvanstrien.xyz/posts/2025/FineWeb-c/scandinavian-content-filtering-fineweb.html
FineWeb2 is a massive multilingual dataset for pre-training language models. Like any web-scale dataset, it contains low-quality content. How can we improve it?
Over the past months, an amazing community of 400+ annotators has been labelling content quality (using Argilla) across 23 languages through the FineWeb-C initiative.
Today, I'm happy to share the first classifier trained on this data.
🔍 What we've built:
- A lightweight classifier that efficiently removes low-quality content
- 90%+ precision demonstrated on Danish & Swedish
- Can process the 43M+ documents in Danish FineWeb2 with minimal compute
🌍 Why this matters: The approach can be reproduced for any of the 23 languages in FineWeb-C ( data-is-better-together/fineweb-c). We can improve training data quality at scale without massive compute resources by starting with community annotations and training small, efficient classifiers.
Want to build a classifier for your language? Check out the full blog post with code examples and implementation details: https://danielvanstrien.xyz/posts/2025/FineWeb-c/scandinavian-content-filtering-fineweb.html
-
- 1 reply
davanstrien
posted
an
update
8 days ago
Post
2075
The
data-is-better-together/fineweb-c dataset is growing!
This week a few more languages have got 1,000 annotations for the educational quality of data from HuggingFaceFW/fineweb-2.
Why should you care?
The quality of pre-training data can have a big impact on the performance of downstream language models trained on that data ( HuggingFaceFW/blogpost-fineweb-v1).
Being able to filter by educational quality is on way of improving the quality of the data you use for training an LLM. Very importantly this approach can also reduce the amount of data needed for pertaining.
Why not use an LLM?
LLMs can be used to annotate educational quality for a subset of data. This data can then be used to train a smaller encoder only model to label the full dataset. However, this may not work well for languages outside of english. This is where fineweb-c (community) comes in.
The community is annotating the educational quality of fineweb2 data. Currently 114 languages have some annotations. These annotations will enable a number of things:
- Evaluate whether an LLM can label the educational quality for texts in that language well
- Directly be used for training quality classifiers
- Help discover other rules and huerisitcs for refining fineweb2 further for different languages.
This week the following languages where done:
Swedish thanks to: @Lauler @AntonVic @ohallstrom @bjarlestam @menbom @Ekgren @apsod
Ukrainian thanks to: @hannayukhymenko @robinhad @realPivo @RabotiahovDmytro @reciprocate
Assamese thanks to: @moyoor97 @Arpanjyoti @nawaf-helmi123 @pahigogoi1 @aelhence @kishorekashyap
Want to learn more: https://huggingface.co/blog/davanstrien/fineweb2-community
Contribute yourself here: data-is-better-together/fineweb-c
This week a few more languages have got 1,000 annotations for the educational quality of data from HuggingFaceFW/fineweb-2.
Why should you care?
The quality of pre-training data can have a big impact on the performance of downstream language models trained on that data ( HuggingFaceFW/blogpost-fineweb-v1).
Being able to filter by educational quality is on way of improving the quality of the data you use for training an LLM. Very importantly this approach can also reduce the amount of data needed for pertaining.
Why not use an LLM?
LLMs can be used to annotate educational quality for a subset of data. This data can then be used to train a smaller encoder only model to label the full dataset. However, this may not work well for languages outside of english. This is where fineweb-c (community) comes in.
The community is annotating the educational quality of fineweb2 data. Currently 114 languages have some annotations. These annotations will enable a number of things:
- Evaluate whether an LLM can label the educational quality for texts in that language well
- Directly be used for training quality classifiers
- Help discover other rules and huerisitcs for refining fineweb2 further for different languages.
This week the following languages where done:
Swedish thanks to: @Lauler @AntonVic @ohallstrom @bjarlestam @menbom @Ekgren @apsod
Ukrainian thanks to: @hannayukhymenko @robinhad @realPivo @RabotiahovDmytro @reciprocate
Assamese thanks to: @moyoor97 @Arpanjyoti @nawaf-helmi123 @pahigogoi1 @aelhence @kishorekashyap
Want to learn more: https://huggingface.co/blog/davanstrien/fineweb2-community
Contribute yourself here: data-is-better-together/fineweb-c
- 1 reply
albertvillanova
posted
an
update
11 days ago
Post
1771
Discover all the improvements in the new version of Lighteval: https://huggingface.co/docs/lighteval/
alielfilali01
posted
an
update
11 days ago
Post
1791
3C3H AraGen Leaderboard welcomes today
deepseek-ai/DeepSeek-V3 and 12 other models (including the late gpt-3.5 💀) to the ranking of best LLMs in Arabic !
Observations:
- DeepSeek-v3 ranked 3rd and only Open model among the top 5 !
- A 14B open model ( Qwen/Qwen2.5-14B-Instruct) outperforms gpt-3.5-turbo-0125 (from last year). This shows how much we came in advancing and supporting Arabic presence within the LLM ecosystem !
- Contrary to what observed in likelihood-acc leaderboards (like OALL/Open-Arabic-LLM-Leaderboard) further finetuned models like maldv/Qwentile2.5-32B-Instruct actually decreased the performance compared to the original model Qwen/Qwen2.5-32B-Instruct.
It's worth to note that the decrease is statiscally insignificant which imply that at best, the out-domain finetuning do not really hurts the model original capabilities acquired during pretraining.
Previous work addressed this (finetuning VS pretraining) but more investigation in this regard is required (any PhDs here ? This could be your question ...)
Check out the latest rankings: inceptionai/AraGen-Leaderboard
Observations:
- DeepSeek-v3 ranked 3rd and only Open model among the top 5 !
- A 14B open model ( Qwen/Qwen2.5-14B-Instruct) outperforms gpt-3.5-turbo-0125 (from last year). This shows how much we came in advancing and supporting Arabic presence within the LLM ecosystem !
- Contrary to what observed in likelihood-acc leaderboards (like OALL/Open-Arabic-LLM-Leaderboard) further finetuned models like maldv/Qwentile2.5-32B-Instruct actually decreased the performance compared to the original model Qwen/Qwen2.5-32B-Instruct.
It's worth to note that the decrease is statiscally insignificant which imply that at best, the out-domain finetuning do not really hurts the model original capabilities acquired during pretraining.
Previous work addressed this (finetuning VS pretraining) but more investigation in this regard is required (any PhDs here ? This could be your question ...)
Check out the latest rankings: inceptionai/AraGen-Leaderboard
alielfilali01
posted
an
update
19 days ago
Post
1850
~75% on the challenging GPQA with only 40M parameters 🔥🥳
GREAT ACHIEVEMENT ! Or is it ?
This new Work, "Data Laundering: Artificially Boosting Benchmark Results through Knowledge Distillation", take out the mystery about many models i personally suspected their results. Speacially on leaderboards other than the english one, Like the Open Arabic LLM Leaderbaord OALL/Open-Arabic-LLM-Leaderboard.
The authors of this work, first started by training a model on the GPQA data, which, unsurprisingly, led to the model achieving 100% performance.
Afterward, they trained what they referred to as a 'legitimate' model on legitimate data (MedMCQA). However, they introduced a distillation loss from the earlier, 'cheated' model.
What they discovered was fascinating: the knowledge of GPQA leaked through this distillation loss, even though the legitimate model was never explicitly trained on GPQA during this stage.
This raises important questions about the careful use of distillation in model training, especially when the training data is opaque. As they demonstrated, it’s apparently possible to (intentionally or unintentionally) leak test data through this method.
Find out more: Data Laundering: Artificially Boosting Benchmark Results through Knowledge Distillation (2412.15255)
GREAT ACHIEVEMENT ! Or is it ?
This new Work, "Data Laundering: Artificially Boosting Benchmark Results through Knowledge Distillation", take out the mystery about many models i personally suspected their results. Speacially on leaderboards other than the english one, Like the Open Arabic LLM Leaderbaord OALL/Open-Arabic-LLM-Leaderboard.
The authors of this work, first started by training a model on the GPQA data, which, unsurprisingly, led to the model achieving 100% performance.
Afterward, they trained what they referred to as a 'legitimate' model on legitimate data (MedMCQA). However, they introduced a distillation loss from the earlier, 'cheated' model.
What they discovered was fascinating: the knowledge of GPQA leaked through this distillation loss, even though the legitimate model was never explicitly trained on GPQA during this stage.
This raises important questions about the careful use of distillation in model training, especially when the training data is opaque. As they demonstrated, it’s apparently possible to (intentionally or unintentionally) leak test data through this method.
Find out more: Data Laundering: Artificially Boosting Benchmark Results through Knowledge Distillation (2412.15255)
- 1 reply
davanstrien
posted
an
update
22 days ago
Post
3166
🇸🇰 Hovorte po slovensky? Help build better AI for Slovak!
We only need 90 more annotations to include Slovak in the next Hugging Face FineWeb2-C dataset ( data-is-better-together/fineweb-c) release!
Your contribution will help create better language models for 5+ million Slovak speakers.
Annotate here: data-is-better-together/fineweb-c.
Read more about why we're doing it: https://huggingface.co/blog/davanstrien/fineweb2-community
We only need 90 more annotations to include Slovak in the next Hugging Face FineWeb2-C dataset ( data-is-better-together/fineweb-c) release!
Your contribution will help create better language models for 5+ million Slovak speakers.
Annotate here: data-is-better-together/fineweb-c.
Read more about why we're doing it: https://huggingface.co/blog/davanstrien/fineweb2-community
-
- 3 replies
davanstrien
posted
an
update
29 days ago
Post
1769
Introducing FineWeb-C 🌐🎓, a community-built dataset for improving language models in ALL languages.
Inspired by FineWeb-Edu the community is labelling the educational quality of texts for many languages.
318 annotators, 32K+ annotations, 12 languages - and growing! 🌍
data-is-better-together/fineweb-c
Inspired by FineWeb-Edu the community is labelling the educational quality of texts for many languages.
318 annotators, 32K+ annotations, 12 languages - and growing! 🌍
data-is-better-together/fineweb-c
versae
authored
a
paper
about 1 month ago
alielfilali01
posted
an
update
about 1 month ago
Post
3437
Unpopular opinion: Open Source takes courage to do !
Not everyone is brave enough to release what they have done (the way they've done it) to the wild to be judged !
It really requires a high level of "knowing wth are you doing" ! It's kind of a super power !
Cheers to the heroes here who see this!
Not everyone is brave enough to release what they have done (the way they've done it) to the wild to be judged !
It really requires a high level of "knowing wth are you doing" ! It's kind of a super power !
Cheers to the heroes here who see this!
-
- 5 replies
alielfilali01
posted
an
update
about 1 month ago
Post
1528
Apparently i forgot to put this here !
Well, this is a bit late but consider given our recent blog a read if you are interested in Evaluation.
You don't have to be into Arabic NLP in order to read it, the main contribution we are introducing is a new evaluation measure for NLG. We made the fisrt application of this measure on Arabic for now and we will be working with colleagues from the community to expand it to other languages.
Blog:
Rethinking LLM Evaluation with 3C3H: AraGen Benchmark and Leaderboard
https://huggingface.co/blog/leaderboard-3c3h-aragen
Space:
inceptionai/AraGen-Leaderboard
Give it a read and let me know your thoughts 🤗
Well, this is a bit late but consider given our recent blog a read if you are interested in Evaluation.
You don't have to be into Arabic NLP in order to read it, the main contribution we are introducing is a new evaluation measure for NLG. We made the fisrt application of this measure on Arabic for now and we will be working with colleagues from the community to expand it to other languages.
Blog:
Rethinking LLM Evaluation with 3C3H: AraGen Benchmark and Leaderboard
https://huggingface.co/blog/leaderboard-3c3h-aragen
Space:
inceptionai/AraGen-Leaderboard
Give it a read and let me know your thoughts 🤗
Post
1292
My latest project is the outcome of the last 2+ years working with TPUs from the amazing TPU Research Cloud (TRC) program and training Encoder-only LMs with the TensorFlow Model Garden library.
👉 Link: https://github.com/stefan-it/model-garden-lms
An overview of some features:
- Cheatsheet for setting-up a TPU VM Pod (with all necessary dependencies) to pretrain LMs with TF Model Garden
- Conversion scripts that convert TF Model Garden weights to Hugging Face Transformers-compatible models
- Supported architectures include BERT, BERT with Token Dropping and TEAMS
I also released BERT-based models pretrained on the great Hugging Face FineWeb and FineWeb-Edu datasets (10BT subset). With more to come!
👉 Model Hub Link: https://huggingface.co/model-garden-lms
If you find these resources useful, please give them a like!
Made from Bavarian Oberland with ❤️ and 🥨.
👉 Link: https://github.com/stefan-it/model-garden-lms
An overview of some features:
- Cheatsheet for setting-up a TPU VM Pod (with all necessary dependencies) to pretrain LMs with TF Model Garden
- Conversion scripts that convert TF Model Garden weights to Hugging Face Transformers-compatible models
- Supported architectures include BERT, BERT with Token Dropping and TEAMS
I also released BERT-based models pretrained on the great Hugging Face FineWeb and FineWeb-Edu datasets (10BT subset). With more to come!
👉 Model Hub Link: https://huggingface.co/model-garden-lms
If you find these resources useful, please give them a like!
Made from Bavarian Oberland with ❤️ and 🥨.