AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is There Any Work That Looks to Extend This Beyond the Math and Symbolic Domain?
The method is based on the GSM8 data set and MAWPS mobs, I guess. These are mathematical word problems. And this first evaluation was mostly looking at how well you can do a reason through some of those. So if you give them tasks which have finding a character inside a long string, like what is the fifth character or something like that, it can do it. If you don't give it a few examples of how to break it down, the models are very bad at being able to do this. The results were much better than anything that we had before.