datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
dolphinDolphin 🐬
https://erichartford.com/dolphin
Dataset details
This dataset is an attempt to replicate the results of Microsoft's Orca
Our dataset consists of:
~1 million of FLANv2 augmented with GPT-4 completions (flan1m-alpaca-uncensored.jsonl)
~3.5 million of FLANv2 augmented with GPT-3.5 completions (flan5m-alpaca-uncensored.jsonl)
We followed the submix and system prompt distribution outlined in the Orca paper. With a few exceptions. We included all 75k of CoT in the FLAN-1m… See the full description on the dataset page: https://huggingface.co/datasets/QuixiAI/dolphin.dolphin-coder
dolphin-coder
This dataset is transformed from https://www.kaggle.com/datasets/erichartford/leetcode-rosetta
it is used to train dolphin-coder model
dolphin-r1
Dolphin R1 🐬
An Apache-2.0 dataset curated by Eric Hartford and Cognitive Computations
Discord: https://discord.gg/cognitivecomputations
Sponsors
Our appreciation for the generous sponsors of Dolphin R1 - Without whom this dataset could not exist.
Dria https://x.com/driaforall - Inference Sponsor (DeepSeek)
Chutes https://x.com/rayon_labs - Inference Sponsor (Flash)
Crusoe Cloud - Compute Sponsor
Andreessen Horowitz - provided the grant that originally launched… See the full description on the dataset page: https://huggingface.co/datasets/QuixiAI/dolphin-r1.QuixiAI-dolphin-distill
Clean QuixiAI/dolphin-distill dataset
This is an unofficial, reformatted version of QuixiAI/dolphin-distill.
It contains mostly English instruction following and conversation datasets.
Major changes:
only kept the longest valid conversation from each row (optional system prompt, followed by alternating user and gpt turns)
duplicate rows removed
URLs, e-mail addresses, phone numbers, API keys and tokens redacted
shuffled and split into chunks
This filtered the original 11,625,521… See the full description on the dataset page: https://huggingface.co/datasets/agentlans/QuixiAI-dolphin-distill.dolphin-r1-deepseek-stratifiedThis is a direct copy of mlabonne/dolphin-r1-deepseek with extra stratification information added in post-processing.
The strata include:
task_type
output_length
response_style
domain
complexity
There has been no validation done on the strata information, use at your own risk.
OpenCoder-LLM_opc-sft-stage1-DolphinLabeled
OpenCoder-LLM SFT DolphinLabeled
Part of the DolphinLabeled series of datasets
Presented by Eric Hartford and Cognitive Computations
The purpose of this dataset is to enable filtering of OpenCoder-LLM SFT dataset.
The original dataset is OpenCoder-LLM/opc-sft-stage1
I have modified the dataset using two scripts.
dedupe.py - removes rows with identical instruction
label.py - adds a "flags" column containing the following boolean values:
"refusal": whether the… See the full description on the dataset page: https://huggingface.co/datasets/QuixiAI/OpenCoder-LLM_opc-sft-stage1-DolphinLabeled.dolphin-sft-v0.1-preferenceThe preference dataset was generated using Mistral-Instruct-v0.1 finetuned on a GPT-4 subset of the Dolphin dataset (16k samples). Link to the model.
Generated responses are labeled as rejected, GPT-4 responses (original Dolphin data) are labeled as accepted.
The motivation was to test out the SPIN paper finetuning methodology.
lm-eval-results-cognitivecomputations-dolphin-2.9-llama3-8b-private
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9-llama3-8b
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9-llama3-8b
The dataset is composed of 62 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 5 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/nyu-dice-lab/lm-eval-results-cognitivecomputations-dolphin-2.9-llama3-8b-private.QuixiAI-dolphin_DatasetDolphin 🐬
https://erichartford.com/dolphin
Dataset details
This dataset is an attempt to replicate the results of Microsoft's Orca
Our dataset consists of:
~1 million of FLANv2 augmented with GPT-4 completions (flan1m-alpaca-uncensored.jsonl)
~3.5 million of FLANv2 augmented with GPT-3.5 completions (flan5m-alpaca-uncensored.jsonl)
We followed the submix and system prompt distribution outlined in the Orca paper. With a few exceptions. We included all 75k of CoT in the FLAN-1m… See the full description on the dataset page: https://huggingface.co/datasets/Maximiliano-Flores-Dev/QuixiAI-dolphin_Dataset.OpenCoder-LLM_opc-sft-stage2-DolphinLabeled
OpenCoder-LLM SFT DolphinLabeled
Part of the DolphinLabeled series of datasets
Presented by Eric Hartford and Cognitive Computations
The purpose of this dataset is to enable filtering of OpenCoder-LLM SFT dataset.
The original dataset is OpenCoder-LLM/opc-sft-stage2
I have modified the dataset using two scripts.
dedupe.py - removes rows with identical instruction
label.py - adds a "flags" column containing the following boolean values:
"refusal": whether the… See the full description on the dataset page: https://huggingface.co/datasets/QuixiAI/OpenCoder-LLM_opc-sft-stage2-DolphinLabeled.dolphin2.9dolphin-r1-backfill-0.0.2allenai_tulu-3-sft-mixture-DolphinLabeled
allenai tulu-3-sft-mixture DolphinLabeled
Part of the DolphinLabeled series of datasets
Presented by Eric Hartford and Cognitive Computations
The purpose of this dataset is to enable filtering of allenai/tulu-3-sft-mixture dataset.
The original dataset is allenai/tulu-3-sft-mixture
I have modified the dataset using two scripts.
dedupe.py - removes rows with identical final message content
label.py - adds a "flags" column containing the following boolean… See the full description on the dataset page: https://huggingface.co/datasets/QuixiAI/allenai_tulu-3-sft-mixture-DolphinLabeled.MaritimeBench
Maritime Bench
本评测集是航运行业首个基于“学科（一级）- 子学科（二级）- 具体考点（三级）”分类体系打造的专业知识评测集，包含1888道客观选择题，覆盖航海、轮机、电子电气员、GMDSS及船员培训等核心领域。评测内容涵盖理论知识、操作技能和行业规范，旨在提升航运领域AI模型的理解与推理能力，确保其在关键知识上的准确性和适应性。同时，本评测集可为航运专业考试、船员培训及资质认证提供自动化测评支持，并优化船舶管理、导航操作、海上通信等场景中的智能问答与决策系统。
MaritimeBench基于行业权威标准，构建了系统、科学的航运知识评测体系，全面评估模型在航海、轮机、电子电气员、GMDSS及船员培训等领域的表现。评测内容深入理论、实践与规范，助力提升AI模型的专业能力。
MaritimeBench评测集亮点
权威性：严格遵循航运行业标准，确保评测科学、实用。
精准分类：采用“学科-子学科-考点”三级框架，评测更具针对性和可扩展性。… See the full description on the dataset page: https://huggingface.co/datasets/Hi-Dolphin/MaritimeBench.dolphin_deA german translation for the cognitivecomputations/dolphin dataset.
Extracted from seedboxventures/multitask_german_examples_32k.
Translation created by seedbox ai for KafkaLM ❤️.
Available for finetuning in hiyouga/LLaMA-Factory.
dolphin-only-gpt-4Dolphin 🐬
https://erichartford.com/dolphin
Dataset details
This dataset is an attempt to replicate the results of Microsoft's Orca
Our dataset consists of:
~1 million of FLANv2 augmented with GPT-4 completions (flan1m-alpaca-uncensored.jsonl)
~3.5 million of FLANv2 augmented with GPT-3.5 completions (flan5m-alpaca-uncensored.jsonl)
We followed the submix and system prompt distribution outlined in the Orca paper. With a few exceptions. We included all 75k of CoT in the FLAN-1m… See the full description on the dataset page: https://huggingface.co/datasets/polymer/dolphin-only-gpt-4.cognitivecomputations__Dolphin3.0-Llama3.2-1B-details
Dataset Card for Evaluation run of cognitivecomputations/Dolphin3.0-Llama3.2-1B
Dataset automatically created during the evaluation run of model cognitivecomputations/Dolphin3.0-Llama3.2-1B
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 5 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__Dolphin3.0-Llama3.2-1B-details.cognitivecomputations__dolphin-2.9.1-yi-1.5-9b-details
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9.1-yi-1.5-9b
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9.1-yi-1.5-9b
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__dolphin-2.9.1-yi-1.5-9b-details.HuggingFaceTB_smoltalk-DolphinLabeled
HuggingFaceTB smoltalk DolphinLabeled
Part of the DolphinLabeled series of datasets
Presented by Eric Hartford and Cognitive Computations
The purpose of this dataset is to enable filtering of HuggingFaceTB/smoltalk dataset.
The original dataset is HuggingFaceTB/smoltalk
I have modified the dataset using two scripts.
dedupe.py - removes rows with identical final message content
label.py - adds a "flags" column containing the following boolean values:
"refusal":… See the full description on the dataset page: https://huggingface.co/datasets/QuixiAI/HuggingFaceTB_smoltalk-DolphinLabeled.cognitivecomputations__dolphin-2.9.3-mistral-7B-32k-details
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9.3-mistral-7B-32k
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9.3-mistral-7B-32k
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__dolphin-2.9.3-mistral-7B-32k-details.NikolaSigmoid__AceMath-1.5B-Instruct-dolphin-r1-200-details
Dataset Card for Evaluation run of NikolaSigmoid/AceMath-1.5B-Instruct-dolphin-r1-200
Dataset automatically created during the evaluation run of model NikolaSigmoid/AceMath-1.5B-Instruct-dolphin-r1-200
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/NikolaSigmoid__AceMath-1.5B-Instruct-dolphin-r1-200-details.cognitivecomputations__dolphin-2.9.3-Yi-1.5-34B-32k-details
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9.3-Yi-1.5-34B-32k
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9.3-Yi-1.5-34B-32k
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__dolphin-2.9.3-Yi-1.5-34B-32k-details.cognitivecomputations__dolphin-2.9.1-llama-3-70b-details
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9.1-llama-3-70b
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9.1-llama-3-70b
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__dolphin-2.9.1-llama-3-70b-details.cognitivecomputations__Dolphin3.0-Qwen2.5-0.5B-details
Dataset Card for Evaluation run of cognitivecomputations/Dolphin3.0-Qwen2.5-0.5B
Dataset automatically created during the evaluation run of model cognitivecomputations/Dolphin3.0-Qwen2.5-0.5B
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__Dolphin3.0-Qwen2.5-0.5B-details.cognitivecomputations__dolphin-2.9.2-qwen2-72b-details
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9.2-qwen2-72b
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9.2-qwen2-72b
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__dolphin-2.9.2-qwen2-72b-details.cognitivecomputations_dolphin-r1-reasoning-deepseekmlabonne_orca-agentinstruct-1M-v1-cleaned-DolphinLabeled
orca-agentinstruct-1M-v1-cleaned DolphinLabeled
Part of the DolphinLabeled series of datasets
Presented by Eric Hartford and Cognitive Computations
The purpose of this dataset is to enable filtering of orca-agentinstruct-1M-v1-cleaned dataset.
The original dataset is mlabonne/orca-agentinstruct-1M-v1-cleaned
(thank you to microsoft and mlabonne)
I have modified the dataset using two scripts.
dedupe.py - removes rows with identical final response.
label.py -… See the full description on the dataset page: https://huggingface.co/datasets/QuixiAI/mlabonne_orca-agentinstruct-1M-v1-cleaned-DolphinLabeled.cognitivecomputations__dolphin-2.9-llama3-8b-details
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9-llama3-8b
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9-llama3-8b
The dataset is composed of 44 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__dolphin-2.9-llama3-8b-details.cognitivecomputations__dolphin-2.9.2-Phi-3-Medium-abliterated-details
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9.2-Phi-3-Medium-abliterated
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9.2-Phi-3-Medium-abliterated
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__dolphin-2.9.2-Phi-3-Medium-abliterated-details.cognitivecomputations__dolphin-2.9.2-qwen2-7b-details
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9.2-qwen2-7b
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9.2-qwen2-7b
The dataset is composed of 38 configuration(s), each one corresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__dolphin-2.9.2-qwen2-7b-details.