README.md

metadata

title: ViTeX-Bench Leaderboard
emoji: 🏆
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.49.1
app_file: app.py
pinned: false
license: apache-2.0
short_description: Public leaderboard for video scene text editing.

🏆 ViTeX-Bench Leaderboard

🌐 Project page  ·  📊 Dataset  ·  🧪 Benchmark code  ·  🤖 Model & Inference code  ·  🏆 Leaderboard

Public ranking for video scene text editing under the 13-metric, three-axis protocol of ViTeX-Bench.

The full thirteen-metric vector is the unit of report. The table is sorted by TextScore = ∛(SeqAcc · CharAcc · TTS), the geometric mean of the three text-correctness primitives. TextScore is a single-axis sort key by design — no cross-axis aggregate is computed, because no axis substitutes for another. SeqAcc = 0 collapses TextScore to zero, the intended semantics for methods that never produce the requested target string.

Submitting

1. Run the official benchmark on the 157-clip frozen evaluation split: bash scripts/run_benchmark.sh <your_method> in the Benchmark code repo.
2. Upload the produced outputs/<method>/eval.json via the Submit tab.
3. The maintainers review the submission. Approved entries appear on the leaderboard.

Companion repos

• 🌐 Project page: https://vitex-bench.github.io/
• 📊 Dataset: https://huggingface.co/datasets/ViTeX-Bench/ViTeX-Dataset
• 🧪 Benchmark code: https://huggingface.co/ViTeX-Bench/ViTeX-Bench
• 🤖 Model & Inference code (ViTeX-Edit-14B): https://huggingface.co/ViTeX-Bench/ViTeX-Edit-14B