jet-ai/ruler-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.
Add 98304/163840/196608 buckets (RULER @38da79d7; byte-identical-to-published-65536 fidelity gate)
Document 98304 bucket provenance
Add 98304 bucket (RULER @38da79d7; byte-identical-to-published-65536 fidelity gate)
Swap NIAH essay haystack to public-domain Project Gutenberg; relicense CC BY-SA 4.0; bundle generation scripts
Add Nemotron-Nano-v3 tokenized RULER (7 lengths up to 262144, 13 tasks, 100 samples each)
initial commit
