SignalTech/AI
Gamer: A Generalist Vision-Language-Action Foundation Model for Long-Horizon Real-Time Game Agent
High-priority alert based on source significance and keyword relevance.
Context
What happened: Researchers introduced Gamer, a generalist vision-language-action agent framework combining perception, memory, and multi-step reasoning, trained to interact in 2D and 3D game environments.
Key numbers: Trained on tens of thousands of hours of device-level keyboard and
Why it matters: The authors report Gamer generalizes across challenging 3D open-world games with efficiency comparable to skilled players and achieves competitive results on conventional 2D game benchmarks, suggesting repetitive input trajectories can support gains from scaling data and learning algorithms.
What changed
Researchers introduced Gamer, a generalist vision-language-action agent framework combining perception, memory, and multi-step reasoning, trained to interact in 2D and 3D game environments.
Why it matters
The authors report Gamer generalizes across challenging 3D open-world games with efficiency comparable to skilled players and achieves competitive results on conventional 2D game benchmarks, suggesting repetitive input trajectories can support gains from scaling data and learning algorithms.