HistGramBench

HistGramBench measures how well vision-language models redraw printed historical mathematical diagrams as TikZ. Each figure comes with an answer key of typed facts: points, lines, circles, incidences, parallels, labels. A model's TikZ is compiled in a sandbox, its drawing is read back as vector geometry, and each key fact is checked on that geometry or, where geometry cannot decide, by judge models. A figure's score is the share of its key facts the drawing recovers.

Code, protocol and reproduction instructions: github.com/VietAlpha/histgrambench.

Access. Access is granted on request until the accompanying paper is published; the dataset will then be open.

At a glance

Scored figures	2,507 from 95 works, printed 1551–1933
Development figures	150 from 5 works
Answer-key facts	114,403 on the scored figures
Contestant answers	30,084 (6 models × 2 tracks × 2,507 figures)
Tracks	A: image only; B: image plus page text

Loading

from datasets import load_dataset

items = load_dataset("VietAlphaLabs/HistGramBench", "items", split="test")
keys = load_dataset("VietAlphaLabs/HistGramBench", "answer_key_facts", split="test")

row = items[0]
row["image"]                      # the figure given to the model
facts = keys.filter(lambda f: f["item_id"] == row["item_id"] and f["in_key"])

Log in with hf auth login once your access request is approved.

Contents

Config	Content
items	figures with image, page text and metadata
answer_key_facts, answer_key_entities, answer_key_records	answer keys and how they were built
answers	contestant outputs with outcome codes
compiled_pdfs, cgx	compiled drawings and the geometry read from them
verdicts, scores	per-fact verdicts and per-figure scores
works, split_groups	source works and the source groups behind the split

Splits: test is the scored set; development was used to fix thresholds before the test run. item_id joins all tables. Column definitions are in SCHEMA.md.

Licence and citing figures

Apache-2.0 (LICENSE). Most figures come from works held by Wikimedia Commons, the Internet Archive, the Göttingen Digitization Center, Numdam and Project Gutenberg. When you reproduce a figure, credit its source work and repository as listed in works.

Citation

@unpublished{tran2026histgrambench,
  title  = {{HistGramBench}: Measuring Structural Recovery When Multimodal Models Redraw Historical Mathematical Diagrams as {TikZ}},
  author = {Tran, Minh Quang and He, Carter and Bloch, Cameron and Huang, Zitai and Vo, Thien Truong and Tran, Trong Nguyen and Chen, Isaac},
  year   = {2026},
  note   = {Manuscript}
}