Dataset Viewer
Auto-converted to Parquet Duplicate
uid
stringlengths
14
28
source
stringclasses
20 values
capability
stringclasses
9 values
arm
stringclasses
1 value
split_role
stringclasses
1 value
question
stringlengths
17
5.18k
answer
stringlengths
1
1.53k
image
dict
figureqa-train-3419-10
figureqa
chart_reasoning
eval
intervention
Is Violet Red the high median? Answer yes or no.
No.
{ "bytes": [ 137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73, 72, 68, 82, 0, 0, 2, 247, 0, 0, 1, 144, 8, 6, 0, 0, 0, 42, 102, 79, 196, 0, 0, 46, 76, 73, 68, ...
figureqa-train-15686-3
figureqa
chart_reasoning
eval
intervention
Is Dark Seafoam the maximum? Answer yes or no.
No.
{ "bytes": [ 137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73, 72, 68, 82, 0, 0, 2, 252, 0, 0, 1, 144, 8, 6, 0, 0, 0, 210, 134, 180, 51, 0, 0, 74, 239, 73, 68, ...
figureqa-train-12282-2
figureqa
chart_reasoning
eval
intervention
Is Turquoise the minimum? Answer yes or no.
No.
{ "bytes": [ 137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73, 72, 68, 82, 0, 0, 2, 20, 0, 0, 1, 144, 8, 6, 0, 0, 0, 246, 179, 249, 182, 0, 0, 58, 72, 73, 68, ...
dvqa-train-14998-12
dvqa
chart_reasoning
eval
intervention
Does the chart contain any negative values? Provide a succinct answer.
No.
{ "bytes": [ 137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73, 72, 68, 82, 0, 0, 1, 192, 0, 0, 1, 192, 8, 6, 0, 0, 0, 53, 37, 184, 115, 0, 0, 72, 52, 73, 68, ...
plotqa-train-1039-48
plotqa
chart_reasoning
eval
intervention
What is the difference between the Under 5 mortality rate (per 1,000 live births) of Under-5(female) in 2000 and that in 2015? Give a very brief answer.
48.6.
{ "bytes": [ 137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73, 72, 68, 82, 0, 0, 4, 14, 0, 0, 2, 138, 8, 6, 0, 0, 0, 189, 47, 25, 180, 0, 0, 114, 157, 73, 68, ...
dvqa-train-345-0
dvqa
chart_reasoning
eval
intervention
How many objects are preferred by more than 6 people in at least one category? Offer a terse response.
Three.
{ "bytes": [ 137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73, 72, 68, 82, 0, 0, 1, 192, 0, 0, 1, 192, 8, 6, 0, 0, 0, 53, 37, 184, 115, 0, 0, 95, 187, 73, 68, ...
plotqa-train-630-82
plotqa
chart_reasoning
eval
intervention
What is the difference between the Amount exported (in US$) of Goods in 2009 and the Amount exported (in US$) of Services in 2013? Your answer should be compact.
-2.48e+09.
{ "bytes": [ 137, 80, 78, 71, 13, 10, 26, 10, 0, 0, 0, 13, 73, 72, 68, 82, 0, 0, 3, 171, 0, 0, 2, 138, 8, 6, 0, 0, 0, 219, 101, 205, 162, 0, 0, 106, 152, 73, 68, ...
figureqa-train-15490-6
figureqa
chart_reasoning
eval
intervention
Is Dark Orchid the low median? Answer yes or no.
Yes.
{"bytes":"iVBORw0KGgoAAAANSUhEUgAAAj4AAAGQCAYAAACuzGqvAAAkk0lEQVR4nO3dCXTU1Rn38ScJARL2CMSENRHZkoACTX(...TRUNCATED)
figureqa-train-15279-4
figureqa
chart_reasoning
eval
intervention
Is Rosy Brown greater than Sky Blue? Answer yes or no.
Yes.
{"bytes":"iVBORw0KGgoAAAANSUhEUgAAAnQAAAGQCAYAAAAnVP3GAAAuTUlEQVR4nO3dB3hUZfr38TshCSWAEiBA6AjSOyI2BA(...TRUNCATED)
plotqa-train-1715-15
plotqa
chart_reasoning
eval
intervention
"What is the difference between the highest and the lowest cost of damage caused due to the depletio(...TRUNCATED)
1.24e+07.
{"bytes":"iVBORw0KGgoAAAANSUhEUgAABUEAAAK8CAYAAADBM37pAAB5hElEQVR4nO3dB5icVdk/4Gc3PQECgRCMdIgJoBQpCt(...TRUNCATED)
End of preview. Expand in Data Studio

BenchAbility Figure 4 -- held-out eval

The 10% candidate-dev half of the same 884,143-row pool the two training mixtures are drawn from, balanced per capability. Split by image, so no picture here appears in either mixture, and both arms are equally blind to it.

capability n
chart_reading / chart_reasoning 250 / 250
table_lookup / table_reasoning 250 / 250
document_qa / document_text_reading 250 / 250
diagram_and_infographic_understanding 250
scene_text_recognition 250
key_information_extraction 168 (all that exists)

Balanced rather than proportional: the pool is 38% chart_reasoning, so a proportional dev set would measure that leaf precisely and the scarce ones not at all -- and the scarce ones are where the two arms differ most. One row per image, because a second question on the same picture is not an independent measurement.

This is an in-domain dev set, not the paper's evaluation. Same 20 sources, held out by image. It answers "did training move this capability at all", which separates a broken run from a real one. It cannot separate "learned the capability" from "learned these 20 datasets" -- that needs the unseen-source set (ChartQAPro, OCRBench v2, DUDE, MME-RealWorld) which is not in this pool.

Score with evaluate.py from the training bundle; measure the base model first, then each checkpoint, and report gain = checkpoint - base per capability.

Downloads last month
56