Spaces:

My-Weird-Prompts
/

Script-Leaderboard

Running

App Files Files Community

Script-Leaderboard / README.md

danielrosehill

Deploy leaderboard page

a98dff5 verified 1 day ago

preview code

Raw History Blame Contribute Delete

2.17 kB

metadata

title: ScriptBench Leaderboard
emoji: 🏆
colorFrom: indigo
colorTo: pink
sdk: static
app_file: index.html
pinned: false
license: cc-by-4.0
short_description: Which model writes My Weird Prompts, and how others scored
datasets:
  - My-Weird-Prompts/mwp-script-results
  - My-Weird-Prompts/mwp-censorship-results
tags:
  - leaderboard
  - podcast
  - long-form-writing
  - censorship

ScriptBench — the My Weird Prompts script-writer benchmark

ScriptBench is our own internal benchmark, open-sourced. It measures one workload: a ~28k-token brief turned into a ~3,800-word, two-host comedic dialogue for text-to-speech. It probably isn't the right benchmark for your use case. Fork it and tailor the task card, rubric and measures to yours.

Related work: PodBench (Tencent, ACL 2026, arXiv 2601.14903; code, MIT) is the closest public benchmark in this space: 800 samples, inputs up to 21k tokens, multi-speaker podcast scripts, deterministic checks plus an LLM judge.

A static page (no build step) that renders the ScriptBench (formerly MWP-Script) benchmark results and the informal evaluation log. At load it fetches, from the public eval repo's main branch:

results/leaderboard.json, built by scripts/build_leaderboard.py after every run;
results/informal-evaluations.json.

So the page updates when those files are pushed; the Space itself only needs redeploying when the page code changes. Source: spaces/leaderboard/, deployed with scripts/deploy_leaderboard_space.py.