nvidia/HelpSteer
HelpSteer: Helpfulness SteerLM Dataset HelpSteer is an open-source Helpfulness Dataset (CC-BY-4.0) that supports aligning models to become more helpful, factually correct and coherent, while being adjustable in terms of the complexity and verbosity of its responses. Leveraging this dataset and SteerLM, we train a Llama 2 70B to reach 7.54 on MT Bench, the highest among models trained on open-source datasets based on MT Bench Leaderboard as of 15 Nov 2023. This model is available on… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/HelpSteer.
| Author / organization | nvidia |
|---|---|
| License (as declared) | cc-by-4.0 |
| Downloads | 3.5K (as reported by Hugging Face) |
| Likes | 259 |
| Created | 15 November 2023 |
| Last updated at source | 18 December 2024 |
Popularity, last 2 days
3,488 now (+95 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Connected in the knowledge graph
- mradermacher/karakuri-lm-8x7b-chat-v0.1-GGUFModel · used to train
- mradermacher/SOLAR-10.7B-Instruct-Forest-DPO-v1-i1-GGUFModel · used to train
- MaziyarPanahi/bagel-dpo-7b-v0.1-Mistral-7B-Instruct-v0.1-GGUFModel · used to train
- QuantFactory/granite-8b-code-instruct-128k-GGUFModel · used to train
- ibm-granite/granite-8b-code-instruct-4k-GGUFModel · used to train
- mradermacher/karakuri-lm-7b-apm-v0.1-i1-GGUFModel · used to train
- mradermacher/karakuri-lm-7b-apm-v0.1-GGUFModel · used to train
- mradermacher/granite-20b-code-instruct-8k-GGUFModel · used to train
- mradermacher/granite-3b-code-instruct-i1-GGUFModel · used to train
- mradermacher/granite-20b-code-instruct-r1.1-i1-GGUFModel · used to train
- Mungert/granite-8b-code-instruct-128k-GGUFModel · used to train
- Dzluck/granite-3b-code-instruct-2k-GGUFModel · used to train
- mradermacher/granite-34b-code-instruct-8k-i1-GGUFModel · used to train
- mradermacher/nontoxic-bagel-34b-v0.2-i1-GGUFModel · used to train
- mradermacher/bagel-dpo-34b-v0.2-GGUFModel · used to train
- mradermacher/bagel-dpo-8x7b-v0.2-GGUFModel · used to train
- Mungert/granite-8b-code-instruct-4k-GGUFModel · used to train
- Felladrin/Minueza-32M-ChatModel · used to train
- Mungert/granite-3b-code-instruct-128k-GGUFModel · used to train
- mradermacher/granite-8b-code-instruct-4k-i1-GGUFModel · used to train
- mradermacher/mistral-7B-forest-dpo-i1-GGUFModel · used to train
- mradermacher/granite-3b-code-instruct-2k-i1-GGUFModel · used to train
- mradermacher/karakuri-lm-8x7b-chat-v0.1-i1-GGUFModel · used to train
- mradermacher/granite-3b-code-instruct-128k-i1-GGUFModel · used to train
- mradermacher/granite-3b-code-instruct-GGUFModel · used to train
- mradermacher/bagel-dpo-34b-v0.2-i1-GGUFModel · used to train
- ibm-granite/granite-20b-code-instruct-r1.1Model · used to train
- mradermacher/nontoxic-bagel-34b-v0.2-GGUFModel · used to train
- tensorblock/granite-8b-code-instruct-128k-GGUFModel · used to train
- jondurbin/bagel-dpo-34b-v0.2Model · used to train
- mradermacher/granite-20b-code-instruct-8k-i1-GGUFModel · used to train
- mradermacher/granite-3b-code-instruct-128k-GGUFModel · used to train
- mradermacher/UNA-34Beagles-32K-bf16-v1-i1-GGUFModel · used to train
- ibm-granite/granite-3b-code-instruct-2k-GGUFModel · used to train
- mradermacher/granite-8b-code-instruct-GGUFModel · used to train
- legraphista/granite-20b-code-instruct-IMat-GGUFModel · used to train
- mradermacher/bagel-8x7b-v0.2-i1-GGUFModel · used to train
- mratsim/MiniMax-M2.5-FP8-INT4-AWQModel · used to train
- NCSOFT/Llama-3-OffsetBias-8BModel · used to train
- mradermacher/bagel-8x7b-v0.2-GGUFModel · used to train
More from nvidia
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.