sha256
TEMPLAR-I-9312aea2fb742fb6_dataset_json_X-Amz-Algorithm_AWS4-HMAC-SHA256_X-Amz-Credential_AKIAVVLlama-3_2-3B-a8e45611bda6f5f6_dataset_json_X-Amz-Algorithm_AWS4-HMAC-SHA256_X-Amz-Credential_AKILlama-3_2-3B-592baaf9919188dc_dataset_json_X-Amz-Algorithm_AWS4-HMAC-SHA256_X-Amz-Credential_AKIMeta-Llama-3_1-8B-3ae2ae2b013b9daf_dataset_json_X-Amz-Algorithm_AWS4-HMAC-SHA256_X-Amz-CredentiaMeta-Llama-3_1-8B-ec852d871c31dd2f_dataset_json_X-Amz-Algorithm_AWS4-HMAC-SHA256_X-Amz-CredentiaLlama-3_2-3B-d1808b9d666dadb7_dataset_json_X-Amz-Algorithm_AWS4-HMAC-SHA256_X-Amz-Credential_AKILlama-3_2-1B-5359a9c275dc22d7_dataset_json_X-Amz-Algorithm_AWS4-HMAC-SHA256_X-Amz-Credential_AKIMeta-Llama-3_1-8B-b8941feabd434281_dataset_json_X-Amz-Algorithm_AWS4-HMAC-SHA256_X-Amz-Credentia
Datasets
All datasets matching “sha256”round-reduced-sha256-learnability
Round-Reduced SHA-256 Learnability: A Controls-Gated Negative Result
TL;DR
A small CNN learns to distinguish round-reduced SHA-256 outputs from
random with ~100% accuracy through 3 rounds, then collapses to
chance at round 4 and stays there through the full 64 rounds — a sharp
learnability cliff, replicated across 5 seeds and 2 dataset sizes. Full
SHA-256 is statistically indistinguishable from random to these probes
at this budget (a bounded null, not a proof). An… See the full description on the dataset page: https://huggingface.co/datasets/bshepp/round-reduced-sha256-learnability.sha256SHA-256-Proyect