They literally "do context", as they do in-context learning. Also, factuality can effectively be solved by chain-of-thought prompting[0] and the use of external

They literally "do context", as they do in-context learning. Also, factuality can effectively be solved by chain-of-thought prompting[0] and the use of external tools[1], also see toolformerzero[2] for a simple example. [0]: https://arxiv.org/abs/2201.11903 https://arxiv.org/abs/2201.11903 [1]: https://arxiv.org/abs/2302.04761 https://arxiv.org/abs/2302.04761 [2]: https://toolformerzero.com/ https://toolformerzero.com/ [2]: https://toolformerzero.com/ https://toolformerzero.com/