keenanpepper/arc-whestbench-p2-d8b-corpus-14k

Phase 2 D8b training corpus (14,048 MLPs) + k3af teacher bake Random He-init MLPs at the ARC WhiteBox Estimation Challenge Phase 2 shape (16 layers x 1024 width, fp32, N(0, 2/n) weights, ReLU after every layer). Weights are NOT stored - regenerate from the seed: import torch W_SEED0 = 770000 def gen_weights(idx): g = torch.Generator(device="cpu").manual_seed(W_SEED0 + idx) W = torch.randn(16, 1024, 1024, generator=g, dtype=torch.float32) return W * (2.0 / 1024) **… See the full description on the dataset page: https://huggingface.co/datasets/keenanpepper/arc-whestbench-p2-d8b-corpus-14k.

<iframe src="https://silytics.si/embed/item/856daaa4-d8e5-4b56-b949-6a38069811a5" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.