New Paper Introduces PAWBench to Test Probabilistic Alignment in Video Models · Digg

/AI

New Paper Introduces PAWBench to Test Probabilistic Alignment in Video Models

Combined views

15.3K

2 posts, first seen 18h ago

78 likes comments

New Paper Introduces PAWBench to Test Probabilistic Alignment in Video Models

Reactions

2 posts

SP
Sayak Paul@RisingSayak 18 hrs ago
One plausible future is not enough. A video model may generate a convincing die roll, yet produce the same few outcomes again and again. Our new paper asks a simple question: Does the model capture not only what can happen, but how often? Introducing PAWBench 🧵
- likes: 78
- replies: 8
- bookmarks: 34
- reposts: 12
SP
Sayak Paul@RisingSayak 18 hrs ago
One plausible future is not enough. A video model may generate a convincing die roll, yet produce the same few outcomes again and again. Our new paper asks a simple question: Does the model capture not only what can happen, but how often? Introducing PAWBench 🧵
- likes: 78
- replies: 8
- bookmarks: 34
- reposts: 12

Combined views

15.3K

2 posts, first seen 18h ago

78 likes

8

34 saves

17 reposts

8 comments

34 saves

17 reposts

Reactions

2 posts

SP
Sayak Paul@RisingSayak 18 hrs ago
One plausible future is not enough. A video model may generate a convincing die roll, yet produce the same few outcomes again and again. Our new paper asks a simple question: Does the model capture not only what can happen, but how often? Introducing PAWBench 🧵
- likes: 78
- replies: 8
- bookmarks: 34
- reposts: 12
SP
Sayak Paul@RisingSayak 18 hrs ago
One plausible future is not enough. A video model may generate a convincing die roll, yet produce the same few outcomes again and again. Our new paper asks a simple question: Does the model capture not only what can happen, but how often? Introducing PAWBench 🧵
- likes: 78
- replies: 8
- bookmarks: 34
- reposts: 12