LogicBench-1K

Digital logic circuit diagrams with topology-level ground-truth graph annotations.

Official repository: github.com/tkcl-research/LogicBench1k

Paper

Stroke-Level Connectivity Verification: Grounding Vision-Language Models Against Topology Hallucination in Diagram Understanding
Abdullah Ibne Hanif Arean, Niamul Hassan Samin, Md Arifur Rahman, Renu Akter Suity, Juena Ahmed Noshin, and Md Ashikur Rahman
International Conference on Document Analysis and Recognition (ICDAR), 2026

Dataset summary

Property	Value
Samples	1,000
Image format	JPEG, 512x512 px
Annotation format	JSON (directed graph per image)
Train / Val / Test	700 / 100 / 200
Gate vocabulary	AND, OR, NOT, NAND, NOR, XOR, XNOR
Total gates / edges	6,174 / 6,423

Dataset structure

LogicBench-1K/
├── images/
├── annotations/
│   └── splits/
├── metadata/
└── schema/

Splits

Split	File	Count	Sample ID range
Test	annotations/splits/test.txt	200	lb1k_00000-lb1k_00199
Train	annotations/splits/train.txt	700	lb1k_00200-lb1k_00899
Validation	annotations/splits/val.txt	100	lb1k_00900-lb1k_00999

Load a sample

import json
from huggingface_hub import hf_hub_download

sample_id = "lb1k_00042"
ann_path = hf_hub_download("TKCL-HF/LogicBench-1K", f"annotations/{sample_id}.json", repo_type="dataset")
img_path = hf_hub_download("TKCL-HF/LogicBench-1K", f"images/{sample_id}.jpg", repo_type="dataset")

with open(ann_path) as f:
    graph = json.load(f)

Citation

If you use LogicBench-1K, please cite:

Stroke-Level Connectivity Verification: Grounding Vision-Language Models Against Topology Hallucination in Diagram Understanding
Abdullah Ibne Hanif Arean, Niamul Hassan Samin, Md Arifur Rahman, Renu Akter Suity, Juena Ahmed Noshin, and Md Ashikur Rahman
International Conference on Document Analysis and Recognition (ICDAR), 2026

@inproceedings{arean2026slcv,
  author    = {{Abdullah Ibne Hanif Arean} and {Niamul Hassan Samin} and {Md Arifur Rahman} and {Renu Akter Suity} and {Juena Ahmed Noshin} and {Md Ashikur Rahman}},
  title     = {Stroke-Level Connectivity Verification: Grounding Vision-Language Models Against Topology Hallucination in Diagram Understanding},
  booktitle = {International Conference on Document Analysis and Recognition (ICDAR)},
  year      = {2026}
}

License

Creative Commons Attribution 4.0 International (CC BY 4.0)