# file: /home/runner/work/FlashSpec/FlashSpec/scripts/article/make_phase6_figures.py
# hypothesis_version: 6.170.13

[0.01, 0.02, 0.03, 0.04, 0.25, 0.5, 0.75, 0.85, 0.9, 0.95, 1.0, 1.05, 1.1, 7.0, 10.5, 90.0, 150, 220, 270, 296, 316, 318, 372, 380, 400, 600, 700, 720, 744, 760, '#0b0b0b', '#1baf7a', '#2a78d6', '#52514e', '#e6e5e1', '#eb6834', '#eda100', '#fcfcfb', '*.json', '--out', '0 (tie)', '</svg>', 'Best arm per prompt*', 'Code edits', 'HumanEval', 'MT-Bench', 'Mixed (30 prompts)', 'Qwen2.5-0.5B drafter', 'Thompson', 'UCB1 bandit', 'UCB1, as measured', '__main__', 'arm_', 'arms', 'at_divergence', 'bandit_thompson', 'bandit_ucb1', 'best_single_arm', 'cases', 'compiled HF', 'composite', 'composite-g4', 'dataset', 'decoding', 'docs', 'dtype', 'end', 'fig7-fp16-ties.svg', 'float16', 'greedy', 'hf_greedy_compiled', 'hf_static', 'humaneval', 'humaneval/', 'humaneval_edit', 'images', 'lessons', 'margin, compiled HF', 'margin_top1_top2', 'measured', 'method', 'metrics', 'middle', 'mixed', 'mode', 'mt_bench', 'n-gram lookup', 'ngram', 'ngram-g2', 'ngram-g4', 'ngram-g6', 'path', 'paths', 'phase6', 'position', 'prompt_oracles', 'qwen', 'qwen-g2', 'qwen-g4', 'qwen-g6', 'single_arm_step_tps', 'start', 'step_ms_alone', 'step_tps_measured', 't0.7', 'utf-8', 'wrote']