Datasets:
sample_id stringclasses 10
values | image imagewidth (px) 1.28k 1.28k | instruction stringclasses 10
values | website stringclasses 10
values | gold_bbox listlengths 4 4 | action_json stringclasses 5
values | previous_actions listlengths 1 19 | task_id stringclasses 10
values |
|---|---|---|---|---|---|---|---|
0714_X_2_1#2 | λ¬λμ μμνλ €κ³ νλλ° λ€λ₯Έ μ¬λλ€μ΄ μ΄λ€ λ¬λνλ₯Ό μ λμ§ λ£κ³ μΆμ΄. Xμμ λ¬λνλ λΆλ€μκ² μ΄λ€ λ¬λνλ₯Ό μ λμ§ λ¬»λ κΈμ μ¬λ €μ€. | X_2 | [
51.5,
619.5,
601.4,
706.8
] | {"operation": "CLICK", "value": null, "target_text": "μ΄λ©μΌ λλ μ¬μ©μ μ΄λ¦"} | [
"GOTO https://x.com/"
] | 0714_X_2_1 | |
0720_zigzag_1#20 | μ¬λ¦μ νΈνκ² μ
μ κ²μμ λ°ν μνΌμ€λ₯Ό μ°Ύμμ£ΌμΈμ. μ§κ·Έμ¬κ·Έμμ λ°ν μνΌμ€λ₯Ό κ²μνκ³ μμμ λΈλ, κ°κ²©μ 5λ§ μ μ΄νλ‘ μ€μ ν΄μ£ΌμΈμ. ꡬ맀 νκΈ°κ° μΆ©λΆνκ³ λ¬΄λ£λ°°μ‘μ΄ κ°λ₯ν μν νλλ₯Ό κ³¨λΌ μμ¬μ μ΄κΈ°μ₯, μΈν λ°©λ²μ νμΈν λ€ μ°ν΄μ£ΌμΈμ | zigzag | [
297.8,
191.4,
519,
600.7
] | {"operation": "CLICK", "value": null, "target_text": null} | [
"GOTO https://zigzag.kr/",
"CLICK",
"CLICK μΉ΄μΉ΄μ€λ‘ 3μ΄ λ§μ μμνκΈ°",
"CLICK",
"TYPE \"[ID]\" into input",
"CLICK λΉλ°λ²νΈ μ
λ ₯",
"CLICK",
"TYPE \"[PASSWORD]\" into input",
"HOTKEY ['Enter']",
"CLICK ν",
"CLICK",
"CLICK",
"TYPE \"λ°ν μνΌμ€\" into input",
"HOTKEY ['Enter']",
"SCROLL down",
"SCROLL up",
... | 0720_zigzag_1 | |
0714_ebsi_1#2 | λμκ² λ§λ μΈν°λ· κ°μ μ μλμ μ°Ύκ³ μΆμ΄. ebsiμμ κ³ λ±νκ΅ 3νλ
λ―Έμ λΆ κ°μ’λ₯Ό κ²μ ν κ°μ₯ μμ λμ€λ κ°μ’λ₯Ό μ°ν΄μ€. | ebsi | [
767.9,
644.2,
824.7,
665.3
] | {"operation": "CLICK", "value": null, "target_text": "λ«κΈ°"} | [
"GOTO https://www.ebsi.co.kr/ebs/pot/poti/main.ebs"
] | 0714_ebsi_1 | |
0714_google_docs_1#2 | λΉνλ―ΌCμ ν¨λ₯μ κ΄ν κΈμ μ κ³ μΆμ΄. κ΅¬κΈ λ
μ€μμ λΉνλ―ΌCμ ν¨λ₯μ κ΄ν κΈμ μ μ΄μ£Όκ³ pdfλ‘ λ€μ΄λ‘λ ν΄μ€. | google_docs | [
677.2,
151.7,
1231.4,
226.2
] | {"operation": "CLICK", "value": null, "target_text": "μ΄λ©μΌ λλ ν΄λμ ν"} | [
"GOTO https://accounts.google.com/v3/signin/identifier?continue=https%3A%2F%2Fdocs.google.com%2F&dsh=S-1306464316%3A1784019949610766&emr=1&followup=https%3A%2F%2Fdocs.google.com%2F&osid=1&passive=1209600&flowName=GlifWebSignIn&flowEntry=ServiceLogin&ifkv=Ac50bxu4zUBnseXWeGwhxf6scCIFMgepP6dAgTttSf-N5tUkLqp3QMfTLehfz... | 0714_google_docs_1 | |
0714_instagram_2_1#2 | μΈμ€νκ·Έλ¨μμ 'kimsseungmini' κ³μ μ μ€λ λ§λμ μ λ§ λ°κ°μ λ€λ λ΄μ©μ λ©μΈμ§λ₯Ό 2μ€λ‘ 보λ΄μ€. | instagram_2 | [
259.1,
347.8,
998.9,
402.3
] | {"operation": "CLICK", "value": null, "target_text": null} | [
"GOTO https://www.instagram.com/"
] | 0714_instagram_2_1 | |
0714_nate_1#2 | μ¬νκ³νμ μ§°λλ° λ©μΌμ μ μ₯ν΄λκ³ μΆμ΄. λ€μ΄νΈμμ λ΄κ²μ°κΈ°λ‘ μ¬νκ³ν λ©μΌμ μμ± ν μ μ‘ν΄μ€.μ λͺ©μ μ¬νκ³νμΌλ‘ | nate | [
873.9,
283,
1109,
310.9
] | {"operation": "CLICK", "value": null, "target_text": null} | [
"GOTO https://www.nate.com/"
] | 0714_nate_1 | |
0714_naver_webtoon_1#2 | λ€μ΄λ² μΉν°μμ 무ν κ²μ ν μμ 5κ° μν κ΄μ¬λ±λ‘ νκΈ° | naver_webtoon | [
1063.4,
17.5,
1101.9,
34.2
] | {"operation": "CLICK", "value": null, "target_text": "λ‘κ·ΈμΈ"} | [
"GOTO https://comic.naver.com/index"
] | 0714_naver_webtoon_1 | |
0714_ohou_se_1#3 | κ±°μ€μμ μ¬μ©ν λ¬κ·Έλ₯Ό μ¬λ €κ³ νλλ° μ΄λ€κ±Έ 골λΌμΌν μ§ λͺ¨λ₯΄κ² μ΄. μ€λμ μ§μμ λ¬κ·Έλ₯Ό κ²μ ν΄μ£Όκ³ λ°©μκ° κ°λ₯ν μ νμΌλ‘ 3κ° μ°ν΄μ€. | ohou_se | [
694,
580,
752.1,
634
] | {"operation": "CLICK", "value": null, "target_text": null} | [
"GOTO https://ohou.se/",
"CLICK λ‘κ·ΈμΈ"
] | 0714_ohou_se_1 | |
0714_pexels_1#2 | pexelsμμ λ°λ€ κ²μ ν λΉκ΅νκ³ λ€μ΄λ‘λ νκΈ° | pexels | [
144.5,
265.7,
631.7,
305.8
] | {"operation": "CLICK", "value": null, "target_text": null} | [
"GOTO https://www.pexels.com/ko-kr/"
] | 0714_pexels_1 | |
0714_pinterest_1#2 | κ·μ¬μ΄ κ³ μμ΄ μ¬μ§μ λ€μ΄λ‘λ νκ³ μΆμ΄. νν°λ μ€νΈμμ κ³ μμ΄λ₯Ό κ²μνκ³ κ°μ₯ 첫 λ²μ§Έμ λμ€λ μ¬μ§μ λ€μ΄λ‘λ ν΄μ€. | pinterest | [
984.5,
22.9,
1093.6,
91.7
] | {"operation": "CLICK", "value": null, "target_text": "λ‘κ·ΈμΈ"} | [
"GOTO https://kr.pinterest.com/"
] | 0714_pinterest_1 |
π°π· Ko-MultimodalMind2Web
An offline Korean web-agent benchmark, built manually by native Korean annotators.
π 549 tasks Β· πΌοΈ 12,905 scored steps Β· π 144 websites Β· βοΈ ~480 annotator-hours Β· π§ Offline evaluation
π Overview
Ko-MultimodalMind2Web is an offline web-agent benchmark for Korean websites. It contains 549 tasks and 12,905 scored action steps across 144 websites, collected from human browser demonstrations. The published task interface follows Multimodal Mind2Web: given a Korean task, the current screenshot, and preceding action history, an agent predicts the next operation, target, and action value. Evaluation reports element accuracy, step success, and task success.
βοΈ Built by Human Annotators
All tasks and reference demonstrations were created manually by Korean human annotators. They designed realistic requests, operated the websites themselves, and recorded browser demonstrations of each task.
| Item | Details |
|---|---|
| π€ Annotators | 2 native Korean speakers |
| β±οΈ Total effort | About 480 annotator-hours |
This is the shared collection and annotation effort behind all three Korean releases in the suite below. The estimate is based on two annotators working four hours per day, five days per week, for approximately three months (about twelve weeks).
The benchmark follows the offline task/trajectory/action organization used by Mind2Web and Multimodal Mind2Web / SeeAct: language instructions and screenshots are paired with next-action targets. The Korean task demonstrations and annotations were collected manually. The released operation inventory, action history, and accepted alternatives are explicitly retained so evaluation is reproducible on this collection.
How the benchmark was constructed and verified
- Human task design and demonstration. Korean annotators created realistic tasks on Korean websites and Korean-language interfaces, then performed the complete tasks themselves. Browser events, screenshots, and target-element information were captured from their actual interactions.
- Step-level annotation. The recorded demonstrations were organized into task-linked steps with an operation, argument/value, target geometry when relevant, and the preceding action history. Both original and concise task-instruction variants are supplied.
- Manual target correction. Native browser dialogs, separate popup windows, operating-system select menus, and dynamically replaced controls were inspected using their screenshots. Missing or incorrectly aligned targets were corrected manually. Selection actions were checked against subsequent screenshots to distinguish the selected value from a transient highlight.
- Alternative-answer review. Equivalent form-submission actions were represented as alternative valid answers. The final release includes 350 manually located alternative CLICK targets, together with template-matched alternatives and valid Enter-key submissions. Operation, value, and location must belong to the same accepted answer candidate.
- Consistency and scoring checks. All 12,905 scored-step IDs are unique; all steps resolve to one of the 549 tasks and agree with task-level step counts. All 7,197 primary target boxes are finite, have positive area, and lie within their screenshot dimensions. Gold-label round-trip checks return 1.0 for element accuracy, step success, and task success using the included scorer.
π Dataset statistics
Recorded operation distribution
| Category | Count | Share |
|---|---|---|
| CLICK | 5,879 | 45.56% |
| TYPE | 1,399 | 10.84% |
| HOTKEY | 444 | 3.44% |
| SCROLL | 5,007 | 38.80% |
| CLEAR | 15 | 0.12% |
| GO_BACK | 134 | 1.04% |
| DRAG | 27 | 0.21% |
The 549 human reference trajectories contain 14,042 recorded steps. Offline
next-action scoring is restricted to the seven interaction operations listed
above. URL navigation (GOTO), waiting (WAIT), and task termination (FINISH)
are excluded from step-level scoring, leaving 12,905 scored interaction steps.
statistics.json includes per-website counts, task categories, trajectory-length
statistics, and alternative-action counts.
π§© KORMo-VL Benchmark Suite
This dataset is part of the KORMo-VL Korean GUI-agent benchmark suite. Each release is built from manually collected Korean data and follows the protocol of an established international benchmark.
| Dataset | Setting | Follows |
|---|---|---|
| Ko-Screenspot | Single-image GUI grounding | ScreenSpot |
| Ko-MultimodalMind2Web | Offline next-action prediction | Multimodal Mind2Web |
| Ko-Online-Mind2Web | Live task completion | Online Mind2Web |
Download and preview
This repository is private. Authenticate with a Hugging Face account that has access to the repository.
from huggingface_hub import snapshot_download
snapshot_download(
"KORMo-VL/Ko-MultimodalMind2Web",
repo_type="dataset",
local_dir="Ko-MultimodalMind2Web",
allow_patterns=["README.md", "statistics.json", "evaluate.py", "multimodal_mind2web_ko.tar.gz"],
token=True,
)
The dataset viewer contains 10 illustrative examples. The complete benchmark is provided in the separate full-data file listed above. These examples are selected from the full data.
from datasets import load_dataset
preview = load_dataset("KORMo-VL/Ko-MultimodalMind2Web", split="test", token=True)
Extract the full data from the downloaded repository directory:
cd Ko-MultimodalMind2Web
tar -xzf multimodal_mind2web_ko.tar.gz
Evaluation
Use the included evaluate.py to report element accuracy, step success rate,
and task success rate on the released scored steps. The reference scorer
supports accepted alternative actions.
Contact
- KyungTae Lim, Professor at KAIST.
ktlim@kaist.ac.kr
Contributor
- Yerin Nam (yerinnam@kaist.ac.kr)
- Minkyung Cho (minkyung.cho@kaist.ac.kr)
- Minkyu Kim (minkyukim@kaist.ac.kr)
- Seungwoo Song (sswoo@kaist.ac.kr)
- Hangyeol Yoo (hgyoo@kaist.ac.kr)
- Gyun Lee (gyunini@kaist.ac.kr)
- KyungTae Lim (ktlim@kaist.ac.kr)
Citation
References for KORMo and the benchmark protocol followed by this release:
@misc{KORMo,
author = {Minjun Kim and Hyeonseok Lim and Hangyeol Yoo and Inho Won and Seungwoo Song and Minkyung Cho and Junghun Yuk and Changsu Choi and Dongjae Shin and Huije Lee and Hoyun Song and Alice Oh and KyungTae Lim},
title = {KORMo: Korean Open Reasoning Model for Everyone},
year = {2025},
publisher = {GitHub},
journal = {Technical Report},
url = {https://arxiv.org/abs/2510.09426}
}
@inproceedings{zheng2024seeact,
title = {GPT-4V(ision) is a Generalist Web Agent, if Grounded},
author = {Boyuan Zheng and Boyu Gou and Jihyung Kil and Huan Sun and Yu Su},
booktitle = {Forty-first International Conference on Machine Learning},
year = {2024},
url = {https://openreview.net/forum?id=piecKJ2DlB}
}
@inproceedings{deng2023mind2web,
title = {Mind2Web: Towards a Generalist Agent for the Web},
author = {Xiang Deng and Yu Gu and Boyuan Zheng and Shijie Chen and Sam Stevens and Boshi Wang and Huan Sun and Yu Su},
booktitle = {Advances in Neural Information Processing Systems},
volume = {36},
pages = {28091--28114},
year = {2023},
publisher = {Curran Associates, Inc.},
url = {https://proceedings.neurips.cc/paper_files/paper/2023/file/5950bf290a1570ea401bf98882128160-Paper-Datasets_and_Benchmarks.pdf}
}
- Downloads last month
- -