AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Performance of Pairwise Causal Discovery Benchmarks
In several types of questions from these benchmarks that were posed to the models, they did really well. And so I think that- if you could look at activations and things like that, when these questions are being evaluated, you might get some hint as to how they're grounded in the model or something. Yeah. Using kind of probing procedures if the latent representations in the model align with some kind of causal model or causal abstractions or causal assumptions is very fresh research. But my personal belief is that that's the type of analysis you want to run to understand exactly how it's reasoning if causally, if that's what its doing.