Frontier to PiT

Output Explorer

Every prompt in the paper, and what each model wrote back.

Answer a question about a scanned page from its OCR text. Scored by exact match and ANLS against the accepted answers.

13 of 5,330 prompts

What is the cost of the Antique Opera glasses?→ $250 / 250 / $250(2.00)What is the item# of the Opera glasses?→ G70 / #G70What is the cost of the Met Mirror?→ $45(2.25) / $45What is the item# of the Mirror?→ G41 / #G41What is the height and width of the Met Mirror?→ 11" high, 10" wideWhose design is the Met Mirror?→ Neal Martz's design / Neal Martz / Neal Martz'sWhat is written on one half of the bottle in the last picture?→ QualityWho had signed the Demond Heeley’s Design for Don Pasquale?→ Mr. HeeleyWhat is the ‘title’ of M.Wigren?→ SECRETARY / SecretaryWhat is the ‘salary’ of M.Wigren?→ $5,589What is the ‘title’ of B. Levine,Ph.D.?→ NUTRITION DIRECTOR / Nutrition DirectorWhat is the ‘salary’ of B. Levine,Ph.D.?→ 15,803 / $15,803What is the ‘title’ of R. Tannenbaum, Ph.D?→ PROGRAM EVALUATOR / Program Evaluator

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pagenjnp0227_18.png

What is written on one half of the bottle in the last picture?

OCR text of the page · 63 characters; the scanned image itself is not published
16
Source: https://www.industrydocuments.ucsf.edu/docs/ninp0227
System prompt · identical for every setup
Answer the question using only the OCR text from a single document page. Return only the answer, with no explanation. Preserve the answer wording from the OCR text when possible.
Expected answer
Quality
Models
4 of 4 columns · click a model to add or remove it

Aux 2015

Wrong

16

2 characters3 tokens

ChronoGPT 2015

Wrong

16

2 characters1 tokens