v5 (launch): save-before-complete ordering, retry guidance on save failure

keep hint only on data-fidelity axis

guide: anonymous OK, drop reward mention

guide: ask for 10+ ratings per person; clear test data restart

v4: per-rater cumulative count restored from results repo on re-entry

v3: bigger guide font, per-axis hints (esp. data fidelity guidance)

v2: guide text, per-axis question wording, big query card, table accordion, timing logs, my-count

fix: gr.Request injection on submit (keyword-only -> positional); E2E tested

validation as warning toast, keep screen state

gradio 6.26 (requirements.txt)

gradio 6.26 (README.md)

chart A/B human eval app + 913 item pairs (part 6)

chart A/B human eval app + 913 item pairs (part 5)

chart A/B human eval app + 913 item pairs (part 4)

chart A/B human eval app + 913 item pairs (part 3)

chart A/B human eval app + 913 item pairs (part 2)

chart A/B human eval app + 913 item pairs

initial commit