waple0820 commited on
Commit
c020483
·
verified ·
1 Parent(s): 7e213a8

docs: clean README — drop references to non-public mirror

Browse files
Files changed (1) hide show
  1. README.md +3 -7
README.md CHANGED
@@ -22,11 +22,11 @@ configs:
22
 
23
  # LexBench-Browser
24
 
25
- Public release of the LexBench-Browser benchmark for evaluating AI browser agents on real-world web tasks.
26
 
27
  ## Dataset Description
28
 
29
- LexBench-Browser evaluates AI agents on multi-step browsing tasks across 50+ Chinese and English websites — e-commerce, social, video, tools/education, finance, and more. This public release contains **220 tasks** that do not gate on user login, suitable for open evaluation. The full dataset (including login-gated tasks) is available in the private mirror.
30
 
31
  ### Dataset Statistics (v3.0, 2026-04-28)
32
 
@@ -143,15 +143,11 @@ huggingface-cli download Lexmount/LexBench-Browser --repo-type dataset --local-d
143
  - **difficulty**: `easy` | `medium` | `hard`.
144
  - **language**: task description language (`zh` for Chinese, `en` for English).
145
  - **website_region**: target website region (`zh` for Chinese sites, `en` for international sites).
146
- - **login_required**: whether the task is login-gated. Almost all public tasks are `false`; a small number of API-style tasks are `true`.
147
  - **robustness_tags**: ordered, deduplicated list across 6 categories / 16 tags (popup interference, sequence complexity, content dynamics, anti-crawl, localization, complex interaction).
148
 
149
  > **Migration note (v2.x → v3.0)**: 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been removed and the per-tier files (`l1.jsonl`, `l2.jsonl`, `l3-api.jsonl`, `l3-security.jsonl`, `tasks.jsonl`) have been replaced by a single `task.jsonl`.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.
150
 
151
- ## Related Resources
152
-
153
- - [Lexmount/LexBench-Browser-Private](https://huggingface.co/datasets/Lexmount/LexBench-Browser-Private) — full dataset including login-gated tasks (377 tasks).
154
-
155
  ## License
156
 
157
  Apache 2.0
 
22
 
23
  # LexBench-Browser
24
 
25
+ 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.
26
 
27
  ## Dataset Description
28
 
29
+ This release contains **220 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.
30
 
31
  ### Dataset Statistics (v3.0, 2026-04-28)
32
 
 
143
  - **difficulty**: `easy` | `medium` | `hard`.
144
  - **language**: task description language (`zh` for Chinese, `en` for English).
145
  - **website_region**: target website region (`zh` for Chinese sites, `en` for international sites).
146
+ - **login_required**: whether the task is login-gated.
147
  - **robustness_tags**: ordered, deduplicated list across 6 categories / 16 tags (popup interference, sequence complexity, content dynamics, anti-crawl, localization, complex interaction).
148
 
149
  > **Migration note (v2.x → v3.0)**: 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.
150
 
 
 
 
 
151
  ## License
152
 
153
  Apache 2.0