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 · pagemxxj0037_2.png

Two cases each of Bucks Full Flavor and Lights have been received and submitted for which determinations.

OCR text of the page · 2,387 characters; the scanned image itself is not published
Equal Tar with 508 Nicotine Reduction:
Preliminary efforts are underway to develop a new blend using
combinations of the current ART blend, 215-ET and all or part of the
currenti B&H blend.
Project BOLD:
An RTD study to evaluate the MUL blend with 508 expanded at RTD levels
of approximately 75, 100 and 130 mm was completed. Results indicate
subjective preference for the low RTD models (90 and 100mm) .
AMBROSIA:
Monadic Menthol POL-0637 cigarettes (18 CR-2898 by weight on cigarette
wrapper) have been made and shipment is scheduled for the week of May 28.
Models have been made utilizing filler oversprayed with CR-2898 and
Aromatek 150 separately for subjective comparison vs. models with wrapper
and adhesive applications. Evaluations are in progress.
Cigarette wrapper coated with CR-2911 (t-BOC-protected CR-2898) was
compared to CR-2898 on finished cigarettes both subjectively and
analytically.
Evaluations indicate the CR-2911 was subjectively comparable
to CR-2898 .
National Starch is attempting to assess the adhesive incorporating
Aromatek-150 with regard to quality issues. Additionally, they will attempt
to formulate adhesives which contain CR vanillin release compounds (CR-2905
& 2910) .
PROJECT GRAIN:
Cambridge cigarettes made from filler sprayed in the Semiworks showed
comparable anethole and vanillin levels for control versus test aftercuts
(sonolated and nonsonolated) with 30% reduction in alcohol.
Since the
production filler control showed a higher than target anethole level, only
the Semiworks Cambridge control is being compared to the test cigarettes.
Initial panel screening by Flavor Development showed no obvious differences
between the control and sonolated test, but differences were found between
the control and nonsonolated test aftercut cigarettes.
Further tests are in
progress .
PROJECT TARGET (BUCKS) :
Subjective evaluations of previously made Bucks prototypes with
current, two modified and no aftercut are continuing. Additional flavor
work aimed at reducing the rod aroma and increasing the Camel-like character
is in progress.
Two cases each of Bucks Full Flavor and Lights have been received and
submitted for SOLVIC spotting determinations. Results were: 0.5 spots per
2, 0010 cigarettes for the Full Flavor and 1.5 spots per 21 000 for the Lights
products.
-2 -
2022156206
Source: https://www.industrydocuments.ucsf.edu/docs/mxxj0037
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
SOLVIC spottingSOLVIC
Models
4 of 4 columns · click a model to add or remove it

Ours

Partial · ANLS 0.50

SOLVIC spotting determinations

30 characters7 tokens

Aux 2015

Wrong

The question asks for the answer to a single document page, using only the OCR text from a single page. The answer should be a single string, with no explanation.

Looking at the OCR text, I see several sentences that are garbled. Let's try to reconstruct the original text and then answer based on the

302 characters64 tokens

PiT-FT 2015

Wrong

<|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|> <|<assistant|>

120 characters64 tokens

ChronoGPT 2015

Wrong

Equal Tar with 508 Nicotine Reduction: Preliminary efforts are underway to develop a new blend using combinations of the current ART blend, 215-ET and all or part of the currenti B&H blend, Project BOLD: Monadic Menthol POL-0637 cigarettes (18

243 characters64 tokens