Error code:   ConfigNamesError
Exception:    FileNotFoundError
Message:      Couldn't find any data file at /src/services/worker/Multilingual-Multimodal-NLP/WebWorld-SFT-10k. Couldn't find 'Multilingual-Multimodal-NLP/WebWorld-SFT-10k' on the Hugging Face Hub either: FileNotFoundError: Unable to find 'hf://datasets/Multilingual-Multimodal-NLP/WebWorld-SFT-10k@f1d37a75d29c888239e29a4495958182e3034924/data/WebWorld-SFT-10k.jsonl' with any supported extension ['.csv', '.tsv', '.json', '.jsonl', '.ndjson', '.parquet', '.geoparquet', '.gpq', '.arrow', '.txt', '.conll', '.conllu', '.tar', '.xml', '.hdf5', '.h5', '.eval', '.lance', '.tsfile', '.blp', '.bmp', '.dib', '.bufr', '.cur', '.pcx', '.dcx', '.dds', '.ps', '.eps', '.fit', '.fits', '.fli', '.flc', '.ftc', '.ftu', '.gbr', '.gif', '.grib', '.png', '.apng', '.jp2', '.j2k', '.jpc', '.jpf', '.jpx', '.j2c', '.icns', '.ico', '.im', '.iim', '.tif', '.tiff', '.jfif', '.jpe', '.jpg', '.jpeg', '.mpg', '.mpeg', '.msp', '.pcd', '.pxr', '.pbm', '.pgm', '.ppm', '.pnm', '.psd', '.bw', '.rgb', '.rgba', '.sgi', '.ras', '.tga', '.icb', '.vda', '.vst', '.webp', '.wmf', '.emf', '.xbm', '.xpm', '.BLP', '.BMP', '.DIB', '.BUFR', '.CUR', '.PCX', '.DCX', '.DDS', '.PS', '.EPS', '.FIT', '.FITS', '.FLI', '.FLC', '.FTC', '.FTU', '.GBR', '.GIF', '.GRIB', '.PNG', '.APNG', '.JP2', '.J2K', '.JPC', '.JPF', '.JPX', '.J2C', '.ICNS', '.ICO', '.IM', '.IIM', '.TIF', '.TIFF', '.JFIF', '.JPE', '.JPG', '.JPEG', '.MPG', '.MPEG', '.MSP', '.PCD', '.PXR', '.PBM', '.PGM', '.PPM', '.PNM', '.PSD', '.BW', '.RGB', '.RGBA', '.SGI', '.RAS', '.TGA', '.ICB', '.VDA', '.VST', '.WEBP', '.WMF', '.EMF', '.XBM', '.XPM', '.aiff', '.au', '.avr', '.caf', '.flac', '.htk', '.svx', '.mat4', '.mat5', '.mpc2k', '.ogg', '.paf', '.pvf', '.raw', '.rf64', '.sd2', '.sds', '.ircam', '.voc', '.w64', '.wav', '.nist', '.wavex', '.wve', '.xi', '.mp3', '.opus', '.3gp', '.3g2', '.avi', '.asf', '.flv', '.mp4', '.mov', '.m4v', '.mkv', '.webm', '.f4v', '.wmv', '.wma', '.ogm', '.mxf', '.nut', '.AIFF', '.AU', '.AVR', '.CAF', '.FLAC', '.HTK', '.SVX', '.MAT4', '.MAT5', '.MPC2K', '.OGG', '.PAF', '.PVF', '.RAW', '.RF64', '.SD2', '.SDS', '.IRCAM', '.VOC', '.W64', '.WAV', '.NIST', '.WAVEX', '.WVE', '.XI', '.MP3', '.OPUS', '.3GP', '.3G2', '.AVI', '.ASF', '.FLV', '.MP4', '.MOV', '.M4V', '.MKV', '.WEBM', '.F4V', '.WMV', '.WMA', '.OGM', '.MXF', '.NUT', '.glb', '.ply', '.stl', '.GLB', '.PLY', '.STL', '.pdf', '.PDF', '.nii', '.NII', '.zip', '.idx', '.manifest', '.txn']
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/dataset/config_names.py", line 67, in compute_config_names_response
                  config_names = get_dataset_config_names(
                      path=dataset,
                      token=hf_token,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 161, in get_dataset_config_names
                  dataset_module = dataset_module_factory(
                      path,
                  ...<4 lines>...
                      **download_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/load.py", line 1211, in dataset_module_factory
                  raise FileNotFoundError(
                  ...<2 lines>...
                  ) from None
              FileNotFoundError: Couldn't find any data file at /src/services/worker/Multilingual-Multimodal-NLP/WebWorld-SFT-10k. Couldn't find 'Multilingual-Multimodal-NLP/WebWorld-SFT-10k' on the Hugging Face Hub either: FileNotFoundError: Unable to find 'hf://datasets/Multilingual-Multimodal-NLP/WebWorld-SFT-10k@f1d37a75d29c888239e29a4495958182e3034924/data/WebWorld-SFT-10k.jsonl' with any supported extension ['.csv', '.tsv', '.json', '.jsonl', '.ndjson', '.parquet', '.geoparquet', '.gpq', '.arrow', '.txt', '.conll', '.conllu', '.tar', '.xml', '.hdf5', '.h5', '.eval', '.lance', '.tsfile', '.blp', '.bmp', '.dib', '.bufr', '.cur', '.pcx', '.dcx', '.dds', '.ps', '.eps', '.fit', '.fits', '.fli', '.flc', '.ftc', '.ftu', '.gbr', '.gif', '.grib', '.png', '.apng', '.jp2', '.j2k', '.jpc', '.jpf', '.jpx', '.j2c', '.icns', '.ico', '.im', '.iim', '.tif', '.tiff', '.jfif', '.jpe', '.jpg', '.jpeg', '.mpg', '.mpeg', '.msp', '.pcd', '.pxr', '.pbm', '.pgm', '.ppm', '.pnm', '.psd', '.bw', '.rgb', '.rgba', '.sgi', '.ras', '.tga', '.icb', '.vda', '.vst', '.webp', '.wmf', '.emf', '.xbm', '.xpm', '.BLP', '.BMP', '.DIB', '.BUFR', '.CUR', '.PCX', '.DCX', '.DDS', '.PS', '.EPS', '.FIT', '.FITS', '.FLI', '.FLC', '.FTC', '.FTU', '.GBR', '.GIF', '.GRIB', '.PNG', '.APNG', '.JP2', '.J2K', '.JPC', '.JPF', '.JPX', '.J2C', '.ICNS', '.ICO', '.IM', '.IIM', '.TIF', '.TIFF', '.JFIF', '.JPE', '.JPG', '.JPEG', '.MPG', '.MPEG', '.MSP', '.PCD', '.PXR', '.PBM', '.PGM', '.PPM', '.PNM', '.PSD', '.BW', '.RGB', '.RGBA', '.SGI', '.RAS', '.TGA', '.ICB', '.VDA', '.VST', '.WEBP', '.WMF', '.EMF', '.XBM', '.XPM', '.aiff', '.au', '.avr', '.caf', '.flac', '.htk', '.svx', '.mat4', '.mat5', '.mpc2k', '.ogg', '.paf', '.pvf', '.raw', '.rf64', '.sd2', '.sds', '.ircam', '.voc', '.w64', '.wav', '.nist', '.wavex', '.wve', '.xi', '.mp3', '.opus', '.3gp', '.3g2', '.avi', '.asf', '.flv', '.mp4', '.mov', '.m4v', '.mkv', '.webm', '.f4v', '.wmv', '.wma', '.ogm', '.mxf', '.nut', '.AIFF', '.AU', '.AVR', '.CAF', '.FLAC', '.HTK', '.SVX', '.MAT4', '.MAT5', '.MPC2K', '.OGG', '.PAF', '.PVF', '.RAW', '.RF64', '.SD2', '.SDS', '.IRCAM', '.VOC', '.W64', '.WAV', '.NIST', '.WAVEX', '.WVE', '.XI', '.MP3', '.OPUS', '.3GP', '.3G2', '.AVI', '.ASF', '.FLV', '.MP4', '.MOV', '.M4V', '.MKV', '.WEBM', '.F4V', '.WMV', '.WMA', '.OGM', '.MXF', '.NUT', '.glb', '.ply', '.stl', '.GLB', '.PLY', '.STL', '.pdf', '.PDF', '.nii', '.NII', '.zip', '.idx', '.manifest', '.txn']

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

WebWorld-SFT-10k

The Browser as a World Model for Self-Improving Web Code — a 10,000-example SFT release from the WebWorld training set.

Each example is a single-turn improvement of an HTML artifact that was accepted by the browser-issued certificate: the browser re-executed the candidate, target progress held, and every previously verified capability was preserved. Rejected trajectories are not in the export.

Why WebWorld

VLM-driven self-improvement of web code has a structural flaw: the model that proposes the repair is the model that judges it. Both judgments live in screenshot space, so a winning candidate is one that wins visually — exactly the wrong criterion when the failures are behavioral. Submit buttons can be inert, games can ignore keyboard input, and a repair can polish one region while silently breaking a previously working control. The loop calls that progress. Under our audit, the closest re-implementation of a naive VLM critique-and-rewrite cycle lifts HTMLBench by only 0.4 points over Raw, more than ten times less than the full WebWorld gate.

The missing piece is a counterparty the VLM cannot fool, and the browser already is one: a deterministic, executable simulator of how an HTML artifact behaves under user actions, and in everything but name a world model for web code. WebWorld is the interface that lets a VLM prior interact with this browser-as-world-model autonomously and decides which interactions become supervision.

Framework

Each round splits hypothesis from proof by construction:

The runtime executes the current artifact and records GUI evidence.
The VLM proposes a critique.
The planner compiles the critique into a typed interaction contract with a target predicate, replay action, preserve set, and evidence path.
The browser re-executes the candidate and issues an acceptance certificate only when target progress holds and every previously verified capability is preserved.
Certified transitions form a quality ratchet, and the SFT export sees only those.

Under matched training, WebWorld-27B improves Raw-27B by +5.3 on HTMLBench-400 and +14.9 on MiniAppBench-Val, reaching the level of frontier systems on interactive HTML generation. Equal-size ablations show browser-backed admission carries the gain: without the certificate, the matched 9B lift nearly disappears.

Test-set integrity

HTMLBench (400 prompts) and MiniAppBench-Val (100 queries) are held-out evaluation. The full 32.8k WebWorld training set includes trajectories seeded from those benchmark prompts (2,000 rows from HTMLBench, 800 rows from MiniAppBench). Those rows have been removed before sampling this release, and every row here has been verified to contain no HTMLBench prompt substring and no MiniAppBench query substring.

Check	Rule	Result
HTMLBench prompt substring match (all 400 prompts)	full normalized prompt not present in any SFT prompt	0 hits
MiniAppBench query substring match (all 100 queries)	full normalized query not present in any SFT prompt	0 hits

Data at a glance


Examples	10,000
Format	ShareGPT `{"conversations":[{"from":"human"…},{"from":"gpt"…}]}`
Prompt length (p50 / p90)	331 / 3,364 chars
File size	400.6 MB

Each row is a certified single-turn improvement pair:

{"conversations":[
  {"from":"human","value":"Make a single HTML file for a simple racing game …"},
  {"from":"gpt","value":"<!DOCTYPE html>\n<html lang=\"en\"> … </html>"}
]}

What is not included

HTMLBench (6 categories, 400 prompts) — held out for evaluation
MiniAppBench-Val (100 queries) — held out for evaluation
Records/trajectories with internal certificate metadata, DOM probes, and file-system paths
Baseline raw_equal32800 — same-size raw baseline used for the main paper comparison
Acceptance-rule ablations (cert_20k, vlm_only_20k, score_gate_20k, no_preserve_20k, no_certificate_20k)
Depth ablations (depth_ge1/2/3/5/8_equal5k)

Citation

Paper: WebWorld: The Browser as a World Model for Self-Improving Web Code. The arXiv identifier will be added here on release.

License

Apache-2.0.

Downloads last month: 51

Total file size:

402 MB

Collection including Multilingual-Multimodal-NLP/WebWorld-SFT-10k

WebWorld

Collection

The browser as a world model for self-improving web code: certificate-verified SFT data, evaluator and models to follow. • 1 item • Updated 4 days ago