datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Nemotron-Cascade-RL-Math
Nemotron-Cascade-RL-Math
Nemotron-Cascade-RL-Math is a diverse and high-quality dataset focused on math reasoning. It serves as the Math RL data for Nemotron-Cascade.
Nemotron-Cascade-RL-MATH contains 14,476 math problems and short answers, covering the data sources from OpenMathReasoning, NuminaMath-CoT, DeepScaleR, AceReason-Math. We conduct data decontamination and filter the sample that has a 9-gram overlap with any test sample in our math benchmarks.
The following are… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/Nemotron-Cascade-RL-Math.cascade_bench
CascadeBench: Do Enterprise Systems Need Learned World Models?
🎉 Accepted to NeurIPS 2026: The Fortieth Annual Conference on Neural Information Processing Systems (Main Track)
A reasoning-focused benchmark for predicting enterprise business-rule cascades, built on synthetic schemas with rule-level attribution of every field change
About
In enterprise systems, the dynamics come from tenant-specific business logic that varies across deployments and changes over time.… See the full description on the dataset page: https://huggingface.co/datasets/ServiceNow-AI/cascade_bench.sci-agent-verification-cascade
Scientific Agent Verification Cascade
Public evaluation fixtures and verified aggregate results for testing whether
scientific claims keep their source, meaning, uncertainty, and verification
requirements as they move between AI agents.
This dataset accompanies the
Scientific Agent Verification Cascade
codebase. Version 0.2.0
contains synthetic evaluation data and aggregate-only results. It contains no
raw hosted-model response, private holdout identifier,
source-record… See the full description on the dataset page: https://huggingface.co/datasets/jang1563/sci-agent-verification-cascade.reasoning-sft-Nemotron-Cascade-SFT-SWE-210K
reasoning-sft-Nemotron-Cascade-SFT-SWE-210K
Converted version of nvidia/Nemotron-Cascade-SFT-SWE, filtered to thinking=True rows with exactly one valid <think>...</think> block.
Format
Each row has three columns:
input — list of dicts (conversation turns with role and content, system messages dropped, last assistant message removed)
response — assistant response string including <think> reasoning block
domain — {category}_{source} (e.g. SWE Repair_SWE-Fixer-Train)… See the full description on the dataset page: https://huggingface.co/datasets/AmanPriyanshu/reasoning-sft-Nemotron-Cascade-SFT-SWE-210K.Nemotron-Cascade-2-RL-reproduction
Nemotron-Cascade-2 RL — Unified Reconstruction Recipe & Schema Sample
⚠️ これは NVIDIA 公式リリースではありません。 Nemotron-Cascade-2（arXiv:2603.19220）の
RL（事後学習）データを、公開済みの Nemotron 系データから再構成するための「レシピ＋統一スキーマ」
パッケージです。同梱の train.jsonl は構造確認用の合成サンプルで、本物の学習データではありません
（各行 meta.synthetic_placeholder = true）。本物は build_cascade2_rl_data.py --mode full で
各 Nemotron データセットを取得して生成します。
Dataset Summary
Cascade 2 の RL は 7 段（IF-RL → Multi-domain RL → MOPD → RLHF → Long-context RL → Code RL →… See the full description on the dataset page: https://huggingface.co/datasets/TeamDelta/Nemotron-Cascade-2-RL-reproduction.