Dataset

Data sourced from the official HuggingFace dataset. This page is aligned to the latest public task metadata from Hugging Face and shows only real static data.

...
Tasks
...
Profiles
...
Rubrics
...
Files
...
With deps
...
File types

Worker Profile Distribution

Distribution of 388 tasks across 15 public persona values from the official HuggingFace dataset.

Source: HuggingFace
# Worker profile Tasks Share
Loading...

Collaboration Types

Each task is tagged with one or more collaboration types. A single task may require multiple abilities.

# Collaboration type Tasks tagged
Loading...

Rubric Evaluation Types

7,399 rubrics across the public evaluation-type values released in the task metadata.

Evaluation type Count Share
Loading...

Task Browser

Browse all tasks from the dataset. Powered by the WORKSPACE_BENCH_API.

ID Worker profile Task description Rubrics Files
Loading...

API Playground

Inspect the public static dataset snapshot and test local session-only extensions via WORKSPACE_BENCH_API.

Add a task

API log

API Reference

// Load the API (already available on this page) const api = window.WORKSPACE_BENCH_API; // Get current data snapshot api.getSnapshot() // 返回公开静态统计、画像分布、协作类型、 rubric 分布和元数据 api.getProfileDistribution() // 返回 [{profile, tasks, share}, ...] api.getCollabTypes() // 返回 [{type, count}, ...] api.getRubricTypes() // 返回 [{type, count, share}, ...] api.getFileDepStats() // 返回 {withDeps, withoutDeps, share} // Get raw task list (paginated) api.getTasks({profile: 'Backend Developer', q: 'security', limit: 10, offset: 0}) // 返回 {tasks: [...], total, limit, offset} // Add a new task in the current browser session only api.addTask({persona, task, collaboration_type, rubric_types, file_count}) // 返回 {id, ...task} // Update benchmark metadata api.updateMeta({totalTasks: 400, totalRubrics: 8000}) // 返回更新后的元数据 // Insert / update a leaderboard row api.upsertRow({agent, harness, model, rubric_pass_rate: 55.2}) // 返回重新排序后的榜单结果
Data source

All statistics on this page are computed from the official ShenYunTzr/Workspace-Bench HuggingFace dataset. The dataset contains 388 task rows with fields such as absolute_id, persona, task, tested_capabilities, rubrics, rubric_types, file_dep_graph, and data_manifest.