Published August 31, 2026 | Version v1

Silicon Sample Benchmark — Tier 1 secondary-4 entry (team_28): raw generation logs, Llama 3.3 70B

Authors/Creators

  • 1. ROR icon Barcelona Supercomputing Center

Description

Complete, unprocessed raw model completions from the Llama 3.3 70B Tier 1 secondary-4 entry submitted to the Silicon Sample Benchmark by team_28. Contains all 99,120 individual generation attempts (9,912 personas × 10 repeats), before parsing or exclusion — none of which failed to parse for this model (0/99,120). Deposited separately from the main code/predictions repository due to GitHub's 100MB file size limit — see K.2 in that repository's registration.md for the full disclosure context.

Files

run_target_llama33_70b_bf16_raw_text.csv

Files (130.3 MB)

Name Size
md5:c1baa5b17827062e38f0aad0b2d2d2b5
130.3 MB Preview Download

Additional details

Related works