datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Nemotron-ClimbMix
ClimbMix Dataset
🚀 Creating the highest-quality pre-training datasets for LLMs 🌟
📄 PAPER
🤗 CLIMBLAB
🤗 CLIMBMIX
🏠 HOMEPAGE
Figure 1: Continuously training a 1B model yields a 2.0% improvement over Llama-3.2-1B, demonstrating a more efficient scaling trend compared to prior models.
Figure 2: Pre-training a 1B model from scratch on ClimbMix shows better scaling effects than training on other datasets.… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-ClimbMix.Nemotron-RL-Agentic-Conversational-Tool-Use-Pivot-v1
Dataset Description:
We created an RL dataset for conversational tool-use by utilizing existing expert tool-use trajectories. We pose each assistant step of the trajectory as a separate behavior cloning problem where the policy model is incentivized to match the tool call choices of the expert model. Each trajectory includes the use of tools for authentication, data lookup, servicing (i.e. booking reservations, changing them, getting discounts, etc), and more across 838 different… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-RL-Agentic-Conversational-Tool-Use-Pivot-v1.Nemotron-RL-Ultra-Training-Blends
Dataset Description:
This dataset provides Reinforcement Learning (RL) and Multi-teacher On-Policy Distillation (MOPD) training-data blends used by the public Nemotron-3-Ultra post-training recipe. The blends are consumed by the NeMo RL training recipes through the NeMo Gym agent framework, in which each prompt is paired with an agent/environment that returns a verifiable or judge-based reward. Each subset is a separate blend; see the recipe for how the blends are used.
The… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-RL-Ultra-Training-Blends.Nemotron-RL-Agentic-SWE-Pivot-v1
Dataset Description:
The SWE-RL dataset provides GitHub issues for training and validating real-world software engineering agents using the OpenHands environment in NeMo Gym. The dataset is a refactored version of the SWE-Gym and R2E-Gym datasets to support the NeMo Gym input format.
This dataset is released as part of NVIDIA NeMo Gym, a framework for building reinforcement learning environments to train large language models. NeMo Gym contains a growing collection of training… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-RL-Agentic-SWE-Pivot-v1.Nemotron-Cascade-2-RL-data
Dataset Description:
The Nemotron-Cascade-2-RL dataset is a curated reinforcement learning (RL) dataset blend used to train Nemotron-Cascade-2-30B-A3B model. It includes instruction-following RL, multi-domain RL, on-policy distillation, and software engineering RL (SWE-RL) data.
This dataset is ready for commercial use.
The dataset contains the following subset:
IF-RL
Contains 45,879 training samples for instruction-following RL. Our curation process mainly… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Cascade-2-RL-data.Nemotron-RL-Instruction-Following-MultiTurnChat-v1
Dataset Description:
The MultiChallenge Dataset is a rigorous benchmark designed to improve large language models in complex multi-turn conversations by explicitly targeting inference memory, instruction retention, version editing, and self-coherence. It employs a unique "model breaking" methodology where tasks are tested against advanced models (Nemotron-Nano-V2 and Qwen3-235B-A22B-Thinking-2507) to expose failure modes. A sample is only accepted into the dataset if the task is… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-RL-Instruction-Following-MultiTurnChat-v1.Nemotron-RLHF-GenRM-v1
Dataset Description:
This dataset is designed to train Generative Reward Models (GenRMs). It leverages reinforcement learning at scale to train accurate and robust GenRMs that generalize better than traditional Bradley-Terry models and reduce the risk of reward hacking.
The dataset is composed of:
Preference data focused on diverse domains
A synthetic safety blend
The data follows a "meta-prompt" structure where the model is instructed to act as an expert evaluation judge. For… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-RLHF-GenRM-v1.NVIDIA-Nemotron-3-Super-120B-A12B-FP8-eval-logs-and-scoresnemotron-nano-eval-logs-and-scoresruler-100-nemotron
RULER-100 — Nemotron-Nano-v3 tokenized
RULER long-context evaluation data, regenerated with the
nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 (instruct) tokenizer so the labeled context
lengths are exact for that model — instead of drifting, as they do when RULER data tokenized for a
different model (e.g. Qwen3) is fed to Nemotron.
What's here
7 context lengths: 4096, 8192, 16384, 32768, 65536, 131072, 262144 (the model's max).
13 RULER tasks: niah_single_1/2/3… See the full description on the dataset page: https://huggingface.co/datasets/jet-ai/ruler-100-nemotron.nemotron3-super-120b-distill
Nemotron-3-Super-120B Self-Distillation Set (code-heavy)
Created by Daniel Rodd / AeVox.Ai. Part of the AeVox Diffusion Drafter project.
10K greedy/sampled completions generated by nvidia/NVIDIA-Nemotron-3-Super-120B-A12B (full reasoning, temp=1.0/top_p=0.95, max_tokens 8192) for aligning a diffusion speculative-decoding drafter (Nemotron-Labs-Diffusion-3B). Prompt mix leans into coding (~75% code, ~15% reasoning/math, ~10% chat).
Used to train:… See the full description on the dataset page: https://huggingface.co/datasets/DrCubix/nemotron3-super-120b-distill.EpistemeAI__Mistral-Nemo-Instruct-12B-Philosophy-Math-details
Dataset Card for Evaluation run of EpistemeAI/Mistral-Nemo-Instruct-12B-Philosophy-Math
Dataset automatically created during the evaluation run of model EpistemeAI/Mistral-Nemo-Instruct-12B-Philosophy-Math
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"… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/EpistemeAI__Mistral-Nemo-Instruct-12B-Philosophy-Math-details.Trappu__Nemo-Picaro-12B-details
Dataset Card for Evaluation run of Trappu/Nemo-Picaro-12B
Dataset automatically created during the evaluation run of model Trappu/Nemo-Picaro-12B
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 pointing to the latest results.
An additional… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/Trappu__Nemo-Picaro-12B-details.nbeerbower__Mistral-Nemo-Prism-12B-v2-details
Dataset Card for Evaluation run of nbeerbower/Mistral-Nemo-Prism-12B-v2
Dataset automatically created during the evaluation run of model nbeerbower/Mistral-Nemo-Prism-12B-v2
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 pointing to the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/nbeerbower__Mistral-Nemo-Prism-12B-v2-details.avemio__GRAG-NEMO-12B-ORPO-HESSIAN-AI-details
Dataset Card for Evaluation run of avemio/GRAG-NEMO-12B-ORPO-HESSIAN-AI
Dataset automatically created during the evaluation run of model avemio/GRAG-NEMO-12B-ORPO-HESSIAN-AI
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 pointing to the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/avemio__GRAG-NEMO-12B-ORPO-HESSIAN-AI-details.nlpguy__Mistral-NeMo-Minitron-Upscale-v1-details
Dataset Card for Evaluation run of nlpguy/Mistral-NeMo-Minitron-Upscale-v1
Dataset automatically created during the evaluation run of model nlpguy/Mistral-NeMo-Minitron-Upscale-v1
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 pointing to… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/nlpguy__Mistral-NeMo-Minitron-Upscale-v1-details.nbeerbower__mistral-nemo-gutenberg-12B-v3-details
Dataset Card for Evaluation run of nbeerbower/mistral-nemo-gutenberg-12B-v3
Dataset automatically created during the evaluation run of model nbeerbower/mistral-nemo-gutenberg-12B-v3
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 pointing to… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/nbeerbower__mistral-nemo-gutenberg-12B-v3-details.migtissera__Tess-3-Mistral-Nemo-12B-details
Dataset Card for Evaluation run of migtissera/Tess-3-Mistral-Nemo-12B
Dataset automatically created during the evaluation run of model migtissera/Tess-3-Mistral-Nemo-12B
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 pointing to the latest… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/migtissera__Tess-3-Mistral-Nemo-12B-details.Nemotron-RL-Agentic-Conversational-Tool-Use-Pivot-v1
Dataset Description:
We created an RL dataset for conversational tool-use by utilizing existing expert tool-use trajectories. We pose each assistant step of the trajectory as a separate behavior cloning problem where the policy model is incentivized to match the tool call choices of the expert model. Each trajectory includes the use of tools for authentication, data lookup, servicing (i.e. booking reservations, changing them, getting discounts, etc), and more across 838… See the full description on the dataset page: https://huggingface.co/datasets/Arsh9210/Nemotron-RL-Agentic-Conversational-Tool-Use-Pivot-v1.Nemotron-RL-Agentic-SWE-Pivot-v1
Dataset Description:
The SWE-RL dataset provides GitHub issues for training and validating real-world software engineering agents using the OpenHands environment in NeMo Gym. The dataset is a refactored version of the SWE-Gym and R2E-Gym datasets to support the NeMo Gym input format.
This dataset is released as part of NVIDIA NeMo Gym, a framework for building reinforcement learning environments to train large language models. NeMo Gym contains a growing collection of training… See the full description on the dataset page: https://huggingface.co/datasets/Bobollinix/Nemotron-RL-Agentic-SWE-Pivot-v1.STXBP1-RAG-Nemotron
🧬⚡ STXBP1-ARIA RAG Database v10.1 - NVIDIA Nemotron Embeddings
The most advanced RAG database for STXBP1 therapeutic research.
A pre-built ChromaDB vector database containing:
571,816 indexed text chunks from ~17,000 curated PubMed Central (PMC) biomedical papers + 165 base editing analysis entries,
(https://huggingface.co/datasets/SkyWhal3/stxbp1-base-editing-sweep),
embedded with NVIDIA's state-of-the-art Llama-Nemotron-Embed-1B-v2 model featuring 2048-dimensional embeddings.
⚡… See the full description on the dataset page: https://huggingface.co/datasets/SkyWhal3/STXBP1-RAG-Nemotron.Nemotron-RL-Agentic-SWE-Pivot-v1
Dataset Description:
The SWE-RL dataset provides GitHub issues for training and validating real-world software engineering agents using the OpenHands environment in NeMo Gym. The dataset is a refactored version of the SWE-Gym and R2E-Gym datasets to support the NeMo Gym input format.
This dataset is released as part of NVIDIA NeMo Gym, a framework for building reinforcement learning environments to train large language models. NeMo Gym contains a growing collection of training… See the full description on the dataset page: https://huggingface.co/datasets/Arsh9210/Nemotron-RL-Agentic-SWE-Pivot-v1.nvidia__Llama-3.1-Nemotron-70B-Instruct-HF-details
Dataset Card for Evaluation run of nvidia/Llama-3.1-Nemotron-70B-Instruct-HF
Dataset automatically created during the evaluation run of model nvidia/Llama-3.1-Nemotron-70B-Instruct-HF
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 pointing… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/nvidia__Llama-3.1-Nemotron-70B-Instruct-HF-details.mistralai__Mistral-Nemo-Instruct-2407-details
Dataset Card for Evaluation run of mistralai/Mistral-Nemo-Instruct-2407
Dataset automatically created during the evaluation run of model mistralai/Mistral-Nemo-Instruct-2407
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 pointing to the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/mistralai__Mistral-Nemo-Instruct-2407-details.nbeerbower__BigKartoffel-mistral-nemo-20B-details
Dataset Card for Evaluation run of nbeerbower/BigKartoffel-mistral-nemo-20B
Dataset automatically created during the evaluation run of model nbeerbower/BigKartoffel-mistral-nemo-20B
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 pointing to… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/nbeerbower__BigKartoffel-mistral-nemo-20B-details.nemocity-worldcognitivecomputations__dolphin-2.9.3-mistral-nemo-12b-details
Dataset Card for Evaluation run of cognitivecomputations/dolphin-2.9.3-mistral-nemo-12b
Dataset automatically created during the evaluation run of model cognitivecomputations/dolphin-2.9.3-mistral-nemo-12b
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"… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/cognitivecomputations__dolphin-2.9.3-mistral-nemo-12b-details.nvidia__Mistral-NeMo-Minitron-8B-Base-details
Dataset Card for Evaluation run of nvidia/Mistral-NeMo-Minitron-8B-Base
Dataset automatically created during the evaluation run of model nvidia/Mistral-NeMo-Minitron-8B-Base
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 pointing to the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/nvidia__Mistral-NeMo-Minitron-8B-Base-details.nemotron_restricted_Final-activationsnvidia__Nemotron-Mini-4B-Instruct-details
Dataset Card for Evaluation run of nvidia/Nemotron-Mini-4B-Instruct
Dataset automatically created during the evaluation run of model nvidia/Nemotron-Mini-4B-Instruct
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 pointing to the latest… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard/nvidia__Nemotron-Mini-4B-Instruct-details.