Update dataset card

Sync dataset card from browseruse-agent-bench repository.

README.md

@@ -1,170 +1,159 @@
 ---
-license: apache-2.0
 language:
-- zh
-- en
-tags:
-- browser-agent
-- web-automation
-- benchmark
-- evaluation
-size_categories:
-- n<1K
+  - zh
+  - en
+license: cc-by-4.0
 task_categories:
-- question-answering
-- text-generation
-configs:
-- config_name: LexBench-Browser
-  data_files:
-  - split: All
-    path: LexBench-Browser/task.jsonl
+  - question-answering
+  - text-generation
+  - reinforcement-learning
+task_ids:
+  - web-agent-benchmark
+pretty_name: LexBench-Browser
+size_categories:
+  - n<1K
+tags:
+  - browser-agent
+  - web-agent
+  - benchmark
+  - evaluation
+  - automation
 ---
 
 # LexBench-Browser
 
-LexBench-Browser is a benchmark for evaluating AI browser agents on real-world web tasks across 50+ Chinese and English websites — e-commerce, social, video, tools/education, finance, and more.
+LexBench-Browser is a public benchmark for evaluating browser agents on real-web workflows. The
+v1.0 snapshot contains 210 no-login tasks across 107 distinct websites, with Chinese and English
+instructions, task-level reference steps, key points, common mistakes, scoring rubrics, and
+robustness tags.
 
-## Dataset Description
+Repository: https://github.com/lexmount/browseruse-agent-bench
 
-This release contains **210 tasks** spanning common multi-step browsing flows. None of the tasks gate on user login, so they can be benchmarked end-to-end without manual sign-in.
+Dataset page: https://huggingface.co/datasets/Lexmount/LexBench-Browser
 
-### Dataset Statistics (v3.1, 2026-04-29)
+Docs: https://docs.bubench.lexmount.io/
 
-| Attribute | Value |
-|-----------|-------|
-| Total tasks | 210 |
-| Version | 3.1 |
-| Languages | Chinese (137), English (73) |
-| `risk_control=true` tasks | 21 |
+## Dataset Summary
 
-#### Task type
+LexBench-Browser is designed for browser-agent engineering:
 
-| Type | Count | Description |
-|------|-------|-------------|
-| T1   | 166   | Information retrieval |
-| T2   | 44    | Website operations |
+- run an agent against real websites
+- compare local and cloud browser backends
+- evaluate task success with a declared judge strategy
+- inspect trajectories and failure modes
+- submit reproducible leaderboard results
 
-#### Reasoning type
+The dataset does not require login for the v1.0 public snapshot.
 
-| Reasoning | Count |
-|-----------|-------|
-| single_step  | 117 |
-| multi_step   | 70  |
-| deep_analysis | 23 |
+## Files
 
-#### Domain Distribution
+```text
+LexBench-Browser/
+|-- data_info.json
+|-- task.jsonl
+|-- task_global.jsonl
+|-- task_lexmount.jsonl
+`-- VERSION_HISTORY.md
+```
 
-| Domain | Count |
-|--------|-------|
-| finance_gaming   | 44 |
-| video_platform   | 42 |
-| tools_education  | 40 |
-| general          | 34 |
-| social_lifestyle | 26 |
-| ecommerce        | 23 |
-| gaming           | 1  |
+Splits:
 
-## Download
+| Split | File | Tasks | Notes |
+| --- | --- | ---: | --- |
+| `All` | `task.jsonl` | 210 | Default public v1.0 split |
+| `global` | `task_global.jsonl` | 92 | Global-region task subset |
+| `lexmount` | `task_lexmount.jsonl` | 118 | Lexmount-region task subset |
 
-### Using Hugging Face Datasets
+## Fields
 
-```python
-from datasets import load_dataset
+Each JSONL row includes:
 
-dataset = load_dataset("Lexmount/LexBench-Browser")
-all_tasks = dataset["All"]
-print(all_tasks[0])
-```
+- `id`: stable task id
+- `query`: user-facing browser task
+- `task_type`: task type label
+- `domain`: domain category
+- `difficulty`: `easy`, `medium`, or `hard`
+- `login_required`: whether login is required
+- `risk_control`: whether the task has risk-control constraints
+- `target_website`: intended website or website family
+- `reasoning_type`: reasoning complexity label
+- `language`: `zh` or `en`
+- `website_region`: expected region/language context
+- `reference_answer`: reference steps, key points, common mistakes, and scoring rubric
+- `score_threshold`: pass threshold
+- `robustness_tags`: practical browser-agent stressors
 
-### For Users in China (Mirror)
+## Label Distribution
 
-```python
-import os
-os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'
+Language:
 
-from datasets import load_dataset
-dataset = load_dataset("Lexmount/LexBench-Browser")
-```
+| Language | Tasks |
+| --- | ---: |
+| `zh` | 137 |
+| `en` | 73 |
 
-Or set the variable before running:
+Reasoning type:
 
-```bash
-export HF_ENDPOINT=https://hf-mirror.com
-python your_script.py
-```
+| Reasoning type | Tasks |
+| --- | ---: |
+| `single_step` | 117 |
+| `multi_step` | 70 |
+| `deep_analysis` | 23 |
 
-### Manual Download
+Domain:
 
-```bash
-pip install huggingface_hub
-huggingface-cli download Lexmount/LexBench-Browser --repo-type dataset --local-dir ./data
-```
+| Domain | Tasks |
+| --- | ---: |
+| `finance_gaming` | 44 |
+| `video_platform` | 42 |
+| `tools_education` | 40 |
+| `general` | 34 |
+| `social_lifestyle` | 26 |
+| `ecommerce` | 23 |
+| `gaming` | 1 |
+
+## Robustness Tags
+
+Tags cover:
 
-For China users:
+- popup interference: `login_popup`, `cookie_consent`, `ad_overlay`
+- sequence complexity: `long_sequence`, `deep_navigation`, `multi_site`
+- content dynamics: `realtime_data`, `lazy_load_scroll`, `iframe_embed`
+- anti-crawl behavior: `captcha_verification`, `anti_bot`, `rate_limiting`
+- localization: `chinese_rendering`, `cross_language`
+- complex interaction: `filter_sort`, `data_extraction`
+
+## Usage
 
 ```bash
-export HF_ENDPOINT=https://hf-mirror.com
-huggingface-cli download Lexmount/LexBench-Browser --repo-type dataset --local-dir ./data
+git clone https://github.com/lexmount/browseruse-agent-bench.git
+cd browseruse-agent-bench
+uv sync --extra browser-use
+uv run bubench run --agent browser-use --data LexBench-Browser --mode first_n --count 3
 ```
 
-## Task Format
-
-```json
-{
-  "id": 1,
-  "query": "Task description",
-  "task_type": "T1",
-  "reasoning_type": "multi_step",
-  "domain": "ecommerce",
-  "difficulty": "medium",
-  "login_required": false,
-  "login_type": "",
-  "risk_control": false,
-  "risk_control_types": [],
-  "target_website": "www.example.com",
-  "language": "zh",
-  "website_region": "zh",
-  "reference_answer": {
-    "steps": ["Step 1", "Step 2"],
-    "key_points": ["Key point 1"],
-    "common_mistakes": ["Common mistake 1"],
-    "scoring": {
-      "total": 100,
-      "items": [{"name": "...", "score": 30, "description": "..."}]
-    }
-  },
-  "score_threshold": 60,
-  "robustness_tags": ["chinese_rendering", "data_extraction"]
-}
-```
+For official leaderboard submissions, follow the repository evaluation protocol and result
+submission docs.
 
-### Field descriptions
+## Licensing
 
-- **task_type**: `T1` (information retrieval) or `T2` (website operations).
-- **reasoning_type**: `single_step` | `multi_step` | `cross_platform` | `deep_analysis`.
-- **domain**: business domain (`ecommerce`, `social_lifestyle`, `video_platform`, `tools_education`, `finance_gaming`, `general`, `gaming`).
-- **difficulty**: `easy` | `medium` | `hard`.
-- **language**: task description language (`zh` for Chinese, `en` for English).
-- **website_region**: target website region (`zh` for Chinese sites, `en` for international sites).
-- **login_required**: whether the task is login-gated (always `false` in this release).
-- **risk_control** / **risk_control_types**: whether the target site applies anti-crawl / risk-control mechanisms (CAPTCHA, slider verification, anti-bot, rate-limiting).
-- **target_website**: a single hostname, or for multi-site tasks, hostnames separated by `+` (e.g. `movie.douban.com + www.imdb.com`).
-- **robustness_tags**: ordered, deduplicated list across 6 categories / 16 tags (popup interference, sequence complexity, content dynamics, anti-crawl, localization, complex interaction).
+The LexBench-Browser benchmark metadata and task definitions are released under CC-BY 4.0.
 
-> **Migration note (v2.x → v3.x)**: the per-record `scenario_tier` field has been removed and the per-tier files (`l1.jsonl`, `l2.jsonl`, `l3-api.jsonl`, `l3-security.jsonl`, `tasks.jsonl`) have been replaced by a single `task.jsonl`. Slice the data with `login_required`, `domain`, or `risk_control` instead.
+This license covers the curated task records, labels, reference steps, scoring rubrics, and
+metadata authored for the benchmark. It does not relicense third-party website content,
+screenshots, traces, marks, page text, or other artifacts collected while running agents.
 
-## License
+Repository code is licensed separately under Apache-2.0.
 
-Apache 2.0
+## Limitations
+
+- Real websites change over time, so tasks may become easier, harder, or temporarily unavailable.
+- Some websites may show region-specific content, anti-bot interstitials, cookie banners, or
+  localized layouts.
+- The public v1.0 snapshot avoids login-required tasks.
+- Automated judge results should be interpreted with the declared judge model, prompt strategy,
+  benchmark version, and browser backend.
 
 ## Citation
 
-```bibtex
-@misc{lexbench-browser-2026,
-  title={LexBench-Browser: A Benchmark for Web Browsing AI Agents},
-  author={Lexmount},
-  year={2026},
-  publisher={Hugging Face},
-  url={https://huggingface.co/datasets/Lexmount/LexBench-Browser}
-}
-```
+Use the repository `CITATION.cff` for citation metadata.