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

Nearby prompts. All 5,330 DocVQA prompts

PromptDocVQA · pageynjf0226_14.png

What type of disease is considered here?

OCR text of the page · 692 characters; the scanned image itself is not published
Evolving Data
Cardiovascular Disease
Demonstrate Improvement in Cardiovascular Risk Factors
GLAI: Lipids effects - Actos vs. Avandia (Q2 2004) - Label claim
IITs: markers of endothelial function, lipids/apolipoproteins,
Delay Progression of Atherosclerosis (surrogate measures)
PERISCOPE: Intravascular ultrasound (Q1 2006) -> Label claim
Electron Beam CT + Carotid Ultrasound (Q1 2006) > Label claim
IITs: Coronary blood flow, Restenosis
Delay or Prevent Cardiovascular Events
PROACTIVE (Q4 2006) -> Label claim
IIT: Stroke prevention
<AUTO>
Confidential - Subject to Protective Order
LLY-HOVENR-00001908
Produced in MDL on 02/27/13
Source: https://www.induposscoooats.ucsf.edu/docs/ynjf0226
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
Cardiovascular DiseaseCardiovascular diseaseCardiovascular
Models
4 of 4 columns · click a model to add or remove it

Ours

Exact match

Cardiovascular Disease

22 characters5 tokens

Aux 2015

Exact match

Cardiovascular Disease

22 characters5 tokens

PiT-FT 2015

Wrong

Empty response.

0 characters

ChronoGPT 2015

Exact match

Cardiovascular Disease

22 characters3 tokens