license: cc-by-4.0
dataset_info:
- config_name: dpo_instruct
features:
- name: chosen
list:
- name: content
dtype: string
- name: role
dtype: string
- name: rejected
list:
- name: content
dtype: string
- name: role
dtype: string
splits:
- name: dolci_flan
num_bytes: 93470441
num_examples: 30009
- name: dolci_instruct_dpo_persona_precise_if
num_bytes: 11675333
num_examples: 4102
- name: dolci_instruct_dpo_precise_if
num_bytes: 56594779
num_examples: 14224
- name: dolci_python_algorithms
num_bytes: 787968960
num_examples: 184361
- name: dolci_sciriff
num_bytes: 43425434
num_examples: 4247
- name: nemotron_code
num_bytes: 244686471
num_examples: 30054
- name: nemotron_instruction_following_chat_v1
num_bytes: 1601476867
num_examples: 70032
- name: nemotron_math
num_bytes: 108803362
num_examples: 21098
- name: nemotron_safety
num_bytes: 100125061
num_examples: 24085
- name: nemotron_stem
num_bytes: 341469602
num_examples: 65738
- name: pleias_rag
num_bytes: 201478603
num_examples: 7891
- name: smol_instruct_rewrite
num_bytes: 2208387
num_examples: 1787
- name: smolrewrite
num_bytes: 10017564
num_examples: 3016
- name: smolsummarize
num_bytes: 26356924
num_examples: 5949
- name: when2call
num_bytes: 36565517
num_examples: 9000
- name: xlam
num_bytes: 224893643
num_examples: 51632
- config_name: sft_instruct
features:
- name: messages
list:
- name: content
dtype: string
- name: role
dtype: string
splits:
- name: dolci_flan
num_bytes: 107784210
num_examples: 3472
- name: dolci_logic_puzzles
num_bytes: 107784210
num_examples: 83472
- name: dolci_instruct_precise_if
num_bytes: 263950453
num_examples: 126271
- name: dolci_python_algorithms
num_bytes: 296864669
num_examples: 186038
- name: dolci_sciriff
num_bytes: 22228226
num_examples: 4460
- name: nemotron_code
num_bytes: 950381906
num_examples: 173047
- name: nemotron_instruction_following_chat_v1
num_bytes: 2922631310
num_examples: 218343
- name: nemotron_math_v2
num_bytes: 484546923
num_examples: 239362
- name: nemotron_posttraining_v2_math_french
num_bytes: 15678092
num_examples: 1000
- name: nemotron_stem
num_bytes: 761734450
num_examples: 348608
- name: pleias_rag
num_bytes: 12851037010
num_examples: 796582
- name: paradocs
num_bytes: 63732438
num_examples: 70000
- name: smol_instruct_rewrite
num_bytes: 5381064
num_examples: 13000
- name: smolrewrite
num_bytes: 24776697
num_examples: 15000
- name: smolsummarize
num_bytes: 80854929
num_examples: 35000
- name: when2call
num_bytes: 15722853
num_examples: 7437
- name: xlam
num_bytes: 116359989
num_examples: 60000
- name: context_qa_hotpot
num_bytes: 474609683
num_examples: 74393
- name: context_qa_tat
num_bytes: 26401532
num_examples: 7195
- name: croissant_aligned_instruct
num_bytes: 4144675
num_examples: 12129
- name: hardcoded_en
num_bytes: 200770
num_examples: 1108
- name: hardcoded_fr
num_bytes: 215218
num_examples: 962
- name: hermes
num_bytes: 16907558
num_examples: 1893
- name: linagora_personas_math
num_bytes: 44022031
num_examples: 16448
- config_name: sft_thinking
features:
- name: messages
list:
- name: content
dtype: string
- name: role
dtype: string
splits:
- name: dolci_think_sft_persona_precise_if
num_bytes: 1274545838
num_examples: 216083
- name: dolci_think_sft_precise_if
num_bytes: 1387191105
num_examples: 118644
- name: nemotron_agentic_toolcalling
num_bytes: 6598244005
num_examples: 293291
- name: nemotron_instruction_following_chat_v1
num_bytes: 1274545838
num_examples: 216083
- name: nemotron_posttrain_v3_math
num_bytes: 4886080343
num_examples: 1251769
- name: nemotron_posttraining_v2_math_french
num_bytes: 1262603542
num_examples: 80000
- name: nemotron_science_v1_mcq
num_bytes: 1724960038
num_examples: 173276
- name: nemotron_code
num_bytes: 5413209846
num_examples: 110796
- name: pleias_rag
num_bytes: 12812113147
num_examples: 796582
- name: opencodereasoning
num_bytes: 16683924673
num_examples: 477868
- name: smoltalk_smolagents_toolcalling
num_bytes: 207000591
num_examples: 9079
- name: synthetic_2_SFT_verified
num_bytes: 2593211389
num_examples: 103254
configs:
- config_name: dpo_instruct
data_files:
- split: dolci_flan
path: dpo_instruct/dolci-flan/*.jsonl
- split: dolci_instruct_dpo_persona_precise_if
path: dpo_instruct/dolci-instruct-dpo-persona-precise-if/*.jsonl
- split: dolci_instruct_dpo_precise_if
path: dpo_instruct/dolci-instruct-dpo-precise-if/*.jsonl
- split: dolci_python_algorithms
path: dpo_instruct/dolci-python-algorithms/*.jsonl
- split: dolci_sciriff
path: dpo_instruct/dolci-sciriff/*.jsonl
- split: nemotron_code
path: dpo_instruct/nemotron-code/*.jsonl
- split: nemotron_instruction_following_chat_v1
path: dpo_instruct/nemotron-instruction-following-chat-v1/*.jsonl
- split: nemotron_math
path: dpo_instruct/nemotron-math/*.jsonl
- split: nemotron_safety
path: dpo_instruct/nemotron-safety/*.jsonl
- split: nemotron_stem
path: dpo_instruct/nemotron-stem/*.jsonl
- split: pleias_rag
path: dpo_instruct/pleias-rag/*.jsonl
- split: smol_instruct_rewrite
path: dpo_instruct/smol-instruct-rewrite/*.jsonl
- split: smolrewrite
path: dpo_instruct/smolrewrite/*.jsonl
- split: smolsummarize
path: dpo_instruct/smolsummarize/*.jsonl
- split: when2call
path: dpo_instruct/when2call/*.jsonl
- split: xlam
path:
- dpo_instruct/xlam-corrupt-json/*.jsonl
- dpo_instruct/xlam-remove-argument/*.jsonl
- dpo_instruct/xlam-remove-tool-call/*.jsonl
- dpo_instruct/xlam-corrupt-arg-llm/*.jsonl
- config_name: sft_instruct
data_files:
- split: dolci_flan
path: sft_instruct/dolci-flan/*.jsonl
- split: dolci_logic_puzzles
path: sft_instruct/dolci-logic-puzzles/*.jsonl
- split: dolci_instruct_precise_if
path: sft_instruct/dolci-instruct-precise-if/*.jsonl
- split: dolci_python_algorithms
path: sft_instruct/dolci-python-algorithms/*.jsonl
- split: dolci_sciriff
path: sft_instruct/dolci-sciriff/*.jsonl
- split: nemotron_code
path: sft_instruct/nemotron-code/*.jsonl
- split: nemotron_instruction_following_chat_v1
path: sft_instruct/nemotron-instruction-following-chat-v1/*.jsonl
- split: nemotron_math_v2
path: sft_instruct/nemotron-math-v2/*.jsonl
- split: nemotron_posttraining_v2_math_french
path: sft_instruct/nemotron-posttraining-v2-math-french/*.jsonl
- split: nemotron_stem
path: sft_instruct/nemotron-stem/*.jsonl
- split: pleias_rag
path: sft_instruct/pleias-rag/*.jsonl
- split: paradocs
path: sft_instruct/paradocs/*.jsonl
- split: smol_instruct_rewrite
path: sft_instruct/smol-instruct-rewrite/*.jsonl
- split: smolrewrite
path: sft_instruct/smolrewrite/*.jsonl
- split: smolsummarize
path: sft_instruct/smolsummarize/*.jsonl
- split: when2call
path: sft_instruct/when2call/*.jsonl
- split: xlam
path: sft_instruct/xlam/*.jsonl
- split: context_qa_hotpot
path: sft_instruct/context-qa-hotpot/*.jsonl
- split: context_qa_tat
path: sft_instruct/context-qa-tat/*.jsonl
- split: croissant_aligned_instruct
path: sft_instruct/croissant-aligned-instruct/*.jsonl
- split: hardcoded_en
path: sft_instruct/hardcoded-en/*.jsonl
- split: hardcoded_fr
path: sft_instruct/hardcoded-fr/*.jsonl
- split: hermes
path: sft_instruct/hermes/*.jsonl
- split: linagora_personas_math
path: sft_instruct/linagora-personas-math/*.jsonl
- config_name: sft_thinking
data_files:
- split: dolci_think_sft_persona_precise_if
path: sft_thinking/dolci-think-sft-persona-precise-if/*.jsonl
- split: dolci_think_sft_precise_if
path: sft_thinking/dolci-think-sft-precise-if/*.jsonl
- split: nemotron_agentic_toolcalling
path: sft_thinking/nemotron-agentic-toolcalling/*.jsonl
- split: nemotron_instruction_following_chat_v1
path: sft_thinking/dolci-think-sft-persona-precise-if/*.jsonl
- split: nemotron_posttrain_v3_math
path: sft_thinking/nemotron-posttrain-v3-math/*.jsonl
- split: nemotron_posttraining_v2_math_french
path: sft_thinking/nemotron-posttraining-v2-math-french/*.jsonl
- split: nemotron_science_v1_mcq
path: sft_thinking/nemotron-science-v1-mcq/*.jsonl
- split: nemotron_code
path: sft_thinking/nemotron-code/*.jsonl
- split: pleias_rag
path: sft_thinking/pleias-rag/*.jsonl
- split: opencodereasoning
path: sft_thinking/opencodereasoning/*.jsonl
- split: smoltalk_smolagents_toolcalling
path: sft_thinking/smoltalk-smolagents-toolcalling/*.jsonl
- split: synthetic_2_SFT_verified
path: sft_thinking/synthetic-2-SFT-verified/*.jsonl
Table of Contents
Dataset Description
The Luciole-PostTraining-Dataset-1.1 is a curated collection of open, instruction-style text data designed for language model post training. It includes a mixture of synthetic and non-synthetic instructions for supervised fine-tuning (SFT) as well as pairs of responses designed for preference alignment (e.g., DPO). With the exception of data for safety alignment, the alignment pairs were generated synthetically with a delta learning approach: all pairs were generated with Qwen3-32B and Qwen3-0.6B and the former were labeled as the accepted responses. Safety data were generated with a mixture of Qwen3-14B , Ministral-3-14B-Instruct, and an interim checkpoint of Luciole-8B-Instruct, after SFT. Pairs were judged with both Ministral-14B-Reasoning and Qwen3-14B. A pair was included only if both models agreed on their labels.
While Luciole-PostTraining-Dataset-1.1 contains some multilingual data, it is primarily English. This version will be followed with versions containing higher levels of multilingual, and especially French, data.
The Luciole PostTraining Dataset was created by the consortium of the OpenLLM France project funded by BPI France as a part of the France 2030 program. Datasets were processed and stored on the GENCI supercomputer Jean Zay, managed by IDRIS.
- License: CC BY-SA 4.0
- Code repository: Luciole-Training
- Paper: coming soon
Curation Rationale
The Luciole-PostTraining-Dataset-1.1 contains only corpora previously published under open licenses or new data generated with open-weights models. It was created in part to facilitate the training of large language models in strict conformance to open-source requirements and European laws on AI development and intellectual property.
Version 1.1 of the dataset is the first step in a larger project of creating a large, open, multilingual post-training dataset, with a particular focus on French.
By sharing our resources openly, we aim to further research on, and development of, multilingual language models.
Bias, Risks, and Limitations
A large portion of data in the Luciole-PostTraining-Dataset-1.1 was generated with third-party, open-weights models which are liable to introduce unwanted linguistic and cultural biases. Efforts to create post-training data targeting languages and cultures in Europe are ongoing and will require an iterative approach.
Due to its role of teaching models what behavior counts as "unsafe", safety alignment pairs are susceptible to contain toxic or dangerous content.
A further limitation of this dataset is that it does not distinguish between variants of different languages. American English and varieties of English spoken in England, for example, are merely labeled as "English". In future work, we hope to focus more on regional linguistic diversity.
Data Subsets
The Luciole-PostTraining-Dataset-1.1 is divided into four subsets:
- [
sft_instruct]: instruction-style data without thinking traces - [
sft_thinking]: instruction-style data with thinking traces - [
dpo_instruct]: instructions with pairs of accepted and rejected responses, without thinking traces - [
dpo_thinking]: instructions with pairs of accepted and rejected responses, with thinking traces (🚧 coming soon)
Sample Metadata
The SFT subsets contain a single [messages] field which provides the content of the sample formatted as a conversation following the HuggingFace chat format.
The DPO subsets contain [chosen] and [rejected] fields, each formatted as a conversation where the content is identical except for the final assistant turns.
Downloading the Data
Available Configurations
The dataset is organized into configurations, each corresponding to a subset of the data used for a particular training stage, e.g. instruction finetuning.
The list of available configurations can be obtained programmatically:
from datasets import get_dataset_config_names
config_names = get_dataset_config_names(
"OpenLLM-France/Luciole-PostTraining-Dataset-1.1"
)
print(config_names)
['dpo_instruct', 'sft_instruct', 'sft_thinking', 'dpo_thinking']
Loading Examples
Load all the samples for a particular configuration, here dpo_instruct:
from datasets import load_dataset
dataset = load_dataset(
"OpenLLM-France/Luciole-PostTraining-Dataset-1.1",
"dpo_instruct"
)
Load the samples for a particular split within a configuration, here smolrewrite in dpo_instruct:
dataset = load_dataset(
"OpenLLM-France/Luciole-PostTraining-Dataset-1.1",
"dpo_instruct",
split="smolrewrite"
)
Details on Data Sources
To preprocess the split datasets, we checked for the presence of names of LLMs and companies (Claude, Amazon, etc.) as well as for Chinese and Russian. Given that most of the data was generated with open source models from Qwen and DeepSeek, there was a preponderance of Chinese and Russian script intermingled with our languages of interest. Samples containing any of these were removed entirely. The prepocessing scripts can be found in this folder of the Luciole-Training repository.
All datasets sourced for different post training phases for Luciole Instruct 1.1 models are listed in the table below, according to data category. Note that for certain datasets we upload our preprocessed version here in its entirety, but only used a randomly selected subsample during training. These are indicated by a weight in cyan, the proportion of the total data which is equal to the number of samples indicated. All of the datasets are in English, except where indicated in violet.
Token Counts
For each subset, the charts below provide a breakdown by token count showing the relative proportions of each data category. The counts for the instruct datasets are based on outputs from the Luciole instruct tokenizer, and thinking datasets from the Luciole thinking tokenizer (🚧 coming soon).
 |
 |
 |
Citation
✍ Paper coming soon!
Acknowledgements
We gratefully acknowledge BPI France for funding the OpenLLM France project under the call "Communs numériques pour l’intelligence artificielle générative" ("Digital commons for generative artificial intelligence").
Processing and storage of the Luciole-PostTraining-Dataset-1.1 was made possible by computing AI and storage resources by GENCI at IDRIS thanks to the grants AD011014561, A0201016189, and AS011016445 on the supercomputer Jean Zay. We gratefully acknowledge support from GENCI and IDRIS and from Stephane Requena (GENCI) and Pierre-François Lavallée (IDRIS) in particular.
The Luciole-PostTraining-Dataset-1.1 was created by members of LINAGORA and OpenLLM-France, including, in alphabetical order:
Akshay Chaturvedi (LINAGORA)
Liam Duignan (CEA List)
Olivier Ferret (CEA List)
Olivier Gouvert (LINAGORA)
Émile Hasard (OpSci)
Julie Hunter (LINAGORA)
Jean-Pierre Lorré (LINAGORA)
Jérôme Louradour (LINAGORA)
Kate Thompson (LINAGORA)
Dev Jerusha Anish Udayan (TALK'R)
Matteo van Ypersele (LINAGORA)
We thank the support team from NVIDIA for their technical and high-level guidance throughout the project, especially: Meriem Bendris, Julia Gusak, Anna Louise Ollerenshaw, Hagit Paz, Christelle Piechurski, Oleg Sudakov.
We are also greatful to the partners of the OpenLLM-France consortium for their valuable input, with particular thanks to (in alphabetical order):
Clément Bénesse (OpSci)
Gabriel Lauzzana (LORIA)
Nathaniël de Leeuw (CEA)
Michel-Marie Maudet (LINAGORA)
We would also like to thank the numerous open data projects that have guided us in the process of creating this dataset, including projects by Pleias, Nvidia, Hugging Face and Allen AI.
Finally, we thank the entire OpenLLM-France community, whose members have helped in diverse ways.