I've developed an empirically-based approach to distinguishing perception from thought and have used this approach to argue that we perceive at least some high-level features. For instance, we sometimes literally perceive the intentions of others.
I've also argued on conceptual grounds that beliefs are necessarily rationally revisable in a certain way. Mental states which lack this capacity aren't beliefs, though they might be some other sort of thought, such as a supposition or a pretense.
In work in progress, I argue that epistemological solipsism is both more resilient and more theoretically important than we might have thought.