Evals / pdf

Stamp and number

5 scenarios, 5 run by an agent. Each run shows its result, time and new tokens.

ScenarioCodexwithout deftCodexwith deft, skill installedCodexwith deft, skill as agents.mdClaude Opus 5.5without deftClaude Opus 5.5with deft, skill installedClaude Codewithout deftClaude Codewith deft
Add page numbersAdd page numbers like "Page 3 of 14" at the bottom of every page of paper.pdf. Save as numbered.pdf.fail53s · 9.8kpass17s · 13.7kpass13s · 10.2kpass80s · 19.4kpass16s · 11.1knot runnot run
Watermark every pageAdd a diagonal CONFIDENTIAL watermark to every page of f1040.pdf and save it as stamped.pdf.fail32s · 8.1kpass17s · 13.8kpass15s · 6.1kpass49s · 21.8kpass11s · 10.2knot runnot run
Merge, number and stamp in one goCombine fw9.pdf and f1040.pdf, number the pages like 1/8, stamp every page with the word DRAFT, and save the result as packet.pdf.pass52s · 14.6kpass23s · 15.1kpass16s · 7.1kpass104s · 21.8kpass16s · 12.4knot runnot run
Chinese page numbers给 f1040.pdf 每页底部加上“第 1 页”这样的页码,保存为 numbered.pdf。pass75s · 8.0kpass15s · 9.9kpass16s · 9.7kpass70s · 20.7kpass52s · 16.3knot runnot run
Watermark with Chinese text给 f1040.pdf 的每一页加上“机密”两个字的斜向水印,保存为 stamped.pdf。pass86s · 14.7kpass16s · 8.3kpass12s · 9.2kpass49s · 14.8kpass25s · 14.7knot runnot run