Jämför metoder

	Multimodal förstärkningsinlärning	Multimodal Transformer
Ämnesområde	Djupinlärning	Djupinlärning
Familj	Machine learning	Machine learning
Ursprungsår	2015–2022	2019–2021
Upphovsperson	Multiple contributors (DeepMind, OpenAI, Google Brain, 2010s–2020s)	Lu et al. (ViLBERT); Radford et al. (CLIP)
Typ	Multimodal deep RL agent	Cross-modal attention-based deep learning model
Ursprungskälla	Reed, S., Zolna, K., Parisotto, E., Colmenarejo, S. G., Novikov, A., Barth-Maron, G., ... & de Freitas, N. (2022). A Generalist Agent. Transactions on Machine Learning Research.	Lu, J., Batra, D., Parikh, D., & Lee, S. (2019). ViLBERT: Pretraining Task-Agnostic Visiolinguistic Representations for Vision-and-Language Tasks. Advances in Neural Information Processing Systems (NeurIPS), 32.
Alias	Multimodal RL, Multi-Sensory Reinforcement Learning, Vision-Language RL, Multi-Input RL	multimodal attention model, cross-modal transformer, vision-language transformer, multi-modal fusion transformer
Närliggande	6	5
Sammanfattning	Multimodal Reinforcement Learning trains agents to make sequential decisions by perceiving and integrating multiple input modalities — such as raw pixels, language instructions, audio, and proprioceptive sensors — simultaneously. Rather than acting on a single data stream, the agent fuses heterogeneous signals into a unified state representation and learns a policy through environmental reward feedback.	A Multimodal Transformer extends the standard Transformer architecture to process and jointly reason over two or more input modalities — most commonly text and images, but also audio, video, or structured data. Cross-modal attention layers allow information from one modality to inform representations in another, enabling tasks such as visual question answering, image captioning, and multimodal sentiment analysis.