JinPLU commited on
Commit
c674d3c
·
verified ·
1 Parent(s): 56a90a7

Clarify navigation across the WRBench release

Browse files
Files changed (1) hide show
  1. README.md +8 -13
README.md CHANGED
@@ -10,25 +10,20 @@ short_description: Persistent-state world model benchmark
10
 
11
  # WRBench
12
 
 
 
13
  WRBench is an open benchmark for camera-controlled video world models and
14
  diagnostic D1-D6 evaluation, introduced in *Current World Models Lack a
15
  Persistent State Core*.
16
 
17
- ## Start Here
18
 
19
- - Leaderboard: https://huggingface.co/spaces/WRBench/wrbench-leaderboard
20
- - Results dataset: https://huggingface.co/datasets/WRBench/wrbench-results
21
- - Benchmark videos: https://huggingface.co/datasets/WRBench/wrbench-videos
22
- - Natural-25 prompts and first frames: https://huggingface.co/datasets/WRBench/wrbench-natural25
23
- - Human annotation labels: https://huggingface.co/datasets/WRBench/wrbench-human-annotations
24
- - Release collection: https://huggingface.co/collections/WRBench/wrbench-current-world-models-lack-a-persistent-state-core
25
 
26
  The paper reports a frozen benchmark table. The Hugging Face results dataset and
27
  leaderboard can be refreshed as new public results are released, so current
28
  model, video, and support counts should be read from the data files.
29
-
30
- ## Links
31
-
32
- - Paper: https://arxiv.org/abs/2606.20545
33
- - Project page: https://jinplu.github.io/WRBench/
34
- - Code: https://github.com/JinPLu/WRBench
 
10
 
11
  # WRBench
12
 
13
+ [Hugging Face Collection](https://huggingface.co/collections/WRBench/wrbench-current-world-models-lack-a-persistent-state-core) · [Project page](https://jinplu.github.io/WRBench/) · [GitHub](https://github.com/JinPLu/WRBench) · [Paper](https://arxiv.org/abs/2606.20545)
14
+
15
  WRBench is an open benchmark for camera-controlled video world models and
16
  diagnostic D1-D6 evaluation, introduced in *Current World Models Lack a
17
  Persistent State Core*.
18
 
19
+ ## Start here
20
 
21
+ - [Leaderboard](https://huggingface.co/spaces/WRBench/wrbench-leaderboard) — Interactive comparison of the current public scores.
22
+ - [Natural-25 inputs](https://huggingface.co/datasets/WRBench/wrbench-natural25) — Prompts, first frames, and paper-input provenance.
23
+ - [Results and addenda](https://huggingface.co/datasets/WRBench/wrbench-results) — Rolling model-level scores and the frozen paper table.
24
+ - [Generated videos and scores](https://huggingface.co/datasets/WRBench/wrbench-videos) — Model outputs with per-video scores and provenance.
25
+ - [Human annotations](https://huggingface.co/datasets/WRBench/wrbench-human-annotations) — Human comparison labels and associated review clips.
 
26
 
27
  The paper reports a frozen benchmark table. The Hugging Face results dataset and
28
  leaderboard can be refreshed as new public results are released, so current
29
  model, video, and support counts should be read from the data files.