SciFigQual-Bench

A Benchmark for Scientific Figure Quality Assessment with Full-Manuscript Context

Overview · Highlights · Files · Rubric · eval1200 · Usage · Citation

---

Gold instances	Public test split	Rubric	SFQ-Agent MAE	Within-1
6,308	eval1200 · 1,200	5 dims · 1–10	0.418	93.4%

ACL · EMNLP · ICML · NeurIPS  |  2020–2025  |  1,144 papers  |  ~355k citing-paragraph records

Figure 1. From isolated-figure evaluation to full-manuscript context: each figure is bound to its caption and citing paragraphs, then scored on five dimensions (1–10).

---

Scientific figure quality in peer review is tri-modal: reviewers cross-check the figure (I), the caption (c), and the citing paragraphs (\mathcal{T}). Natural IQA, AIGC alignment, and chart QA typically score isolated visuals without manuscript evidence.

SciFigQual-Bench binds each published CS figure to full-manuscript context and scores five orthogonal dimensions on a unified 1–10 scale.

This Hugging Face release is the gold dataset: 6,308 expert-rated instances, PNG crops, aggregated human means, and the fixed public test split eval1200.
Evaluation code (Direct / Sidecar / SFQ-Agent) is in the GitHub repository.

---

Highlights

• Full-manuscript context — each instance is ((I, c, \mathcal{T}, m)): figure image, caption, index-driven citing paragraphs, and metadata.
• Large-scale gold — 7,609 curated figures from 1,144 papers; 6,308 exportable expert-rated instances in this package (20,166 raw ratings).
• Top CS venues — ACL, EMNLP, ICML, NeurIPS (2020–2025).
• Five-dimensional rubric — VC, SL, CC, CTX, MR with L1 evidence gating.
• Fixed public split — eval1200 is the paper’s 1,200-figure holdout (254 papers, deterministic paper-level sampling).
• Strong judge baseline — SFQ-Agent + GPT-5.6-Sol (F3) reaches MAE 0.418 and 93.4% within-1 vs. human gold.

---

Overview

SciFigQual-Bench closes the manuscript-context gap for published CS conference figures:

1. Acquire & curate PDFs from four top venues (62,694 raw → 7,609 clean figures).
2. Bind context via PDF index patterns (not abstract-only heuristics).
3. Annotate with a calibrated five-dimensional rubric and release eval1200.

Figure 2. Corpus collection → structure-aware extraction → context binding → five-dimensional rubric → expert validation.

Benchmark at a glance

Raw PDF corpus	62,694
Qualified papers	1,144
Curated figures	7,609
Human gold (this release)	6,308
Raw annotation records	20,166
Public test split	eval1200 · 1,200 figures · 254 papers
Venues / years	ACL, EMNLP, ICML, NeurIPS · 2020–2025
Mean overall human score	8.05 (rated subset)

ACL	EMNLP	ICML	NeurIPS
1,375	1,551	1,595	1,787

_{Gold instances by venue}

Figure 4. Venue–topic radar, construction funnel, score distribution, domain counts, temporal coverage, and per-dimension means.

Takeaways (n=6,308): funnel yield (\eta \approx 12.1%) from strict curation; CC is the weakest axis (mean 7.42) vs. SL (8.58); coverage peaks in 2024–2025.

---

Files

SciFigQual-Bench/
├── figures.jsonl                 6,308 gold instances
├── human_means.csv               aggregated VC / SL / CC / CTX / MR / overall
├── images/                       6,308 PNG crops
├── splits/
│   ├── eval1200.jsonl            public test manifest
│   └── eval1200_meta.json        sampling protocol
├── corpus_stats.json             construction funnel
└── docs/figures/                 paper Figs. 1–4 (shown on this card)

File	What you get
figures.jsonl	One JSON object per figure: caption, citing text, per-rater annotations, human means
human_means.csv	Spreadsheet-friendly gold table
images/*.png	Figure crops; path stored in image
splits/eval1200.jsonl	Paper-reported public test split
splits/eval1200_meta.json	Deterministic selection metadata

Join keys: figure_id  or  (paper_id, fig_index)

Source PDFs and raw crawls are not redistributed.

Instance schema _{(click to expand)}

Field	Type	Description
figure_id	string	Unique id, e.g. 0001_2020.acl-main.1__fig0001
paper_id	string	Paper identifier in this release
fig_index	int	Figure index in the source PDF
venue / year	string / int	ACL, EMNLP, ICML, or NeurIPS · 2020–2025
title	string	Paper title
caption	string	Figure caption (may be empty)
section	string	Nearby section heading
context_texts	list[str]	Index-resolved citing paragraphs (\mathcal{T})
context_count	int	Number of citing paragraphs
image	string	Relative path under images/
annotator_count	int	Number of human ratings aggregated
human_*	float / null	Mean scores after L1 gating
annotations	list[object]	Per-rater scores, summaries, and rationales

Gold scores are the mean over annotations[]. Null means the dimension was gated out (no caption → no CC; no citing text → no CTX).

---

Five-Dimensional Rubric

Scores on a 1–10 Likert scale. Overall is a gated mean over available dimensions.

	Dimension	Measures
VC	Visual Clarity	Readability, resolution, contrast, label legibility
SL	Structure & Layout	Panel organization, alignment, visual hierarchy
CC	Caption Consistency	Caption faithfully describes visible content
CTX	Context Consistency	Figure supports claims in citing paragraphs
MR	Misleading Risk	Truncated axes, missing baselines, deceptive encodings

L1 gating. CC is null without a caption; CTX is null without citing text. Instances lacking both are excluded from scoring.

---

eval1200

The paper’s fixed public test split — 1,200 figures, 254 papers, paper-level all-or-nothing selection (no randomness).

Figures	Papers	With citing context	ACL	EMNLP	ICML	NeurIPS
1,200	254	625 (52.1%)	266	294	294	346

On this split, SFQ-Agent (F3, GPT-5.6-Sol) reports MAE 0.418 and 93.4% within-1 agreement vs. human gold. Prompts and judge protocols: code repository.

Figure 3. L1 gating → parallel vision / language evidence → cross-modal judge → deterministic Runner aggregation.

---

Usage

Download

huggingface-cli download haihanlamu/SciFigQual-Bench \
  --repo-type dataset \
  --local-dir SciFigQual-Bench

Load gold + eval1200

import json
from pathlib import Path

root = Path("SciFigQual-Bench")

with (root / "figures.jsonl").open(encoding="utf-8") as f:
    figures = [json.loads(line) for line in f]

eval_ids = {
    (row["paper_id"], row["fig_index"])
    for line in (root / "splits" / "eval1200.jsonl").open(encoding="utf-8")
    for row in [json.loads(line)]
}

eval1200 = [x for x in figures if (x["paper_id"], x["fig_index"]) in eval_ids]
print(len(figures), len(eval1200))  # 6308  1200

Scoring scripts and the 29-run experiment matrix: github.com/FrankDengAI/SciFigQual-Bench.

---

License

Released under the MIT License. Figure crops and excerpted manuscript text remain subject to the original venue / publisher terms. This package is not a redistribution of full PDFs.

---

Citation

If you use SciFigQual-Bench, please cite:

@article{deng2026scifigqual,
  title={SciFigQual-Bench: A Benchmark for Scientific Figure Quality Assessment with Full-Manuscript Context},
  author={Deng, Zihan and Xu, Chuanzhi and Liang, Huiqi and Li, Haoyang and Zhong, Xiaozhen and Yu, Lequan},
  journal={arXiv preprint arXiv:2607.27084},
  year={2026}
}

Paper: arXiv:2607.27084