Filter by layer, neuron, activating/important tokens, explanation(?), similarity to specific neuron.
there are many possibilities and all should be pretty easy to implement once the main structure is set up.
Since potentially many neurons will be returned, some form of pagination would likely be required.
This could merge both current neuron search methods, but only if it is equally easy to do both.
In the case of the neuron picker, this likely requires some dedicated UI for that functionality.
It should be possible to pick a random neuron in the list.
The GitHub issue list is probably a good source of inspiration.
This could definitely be useful across models, which could be implemented simply by making the model an implicit filter.
Filter by layer, neuron, activating/important tokens, explanation(?), similarity to specific neuron. there are many possibilities and all should be pretty easy to implement once the main structure is set up. Since potentially many neurons will be returned, some form of pagination would likely be required.
This could merge both current neuron search methods, but only if it is equally easy to do both. In the case of the neuron picker, this likely requires some dedicated UI for that functionality.
It should be possible to pick a random neuron in the list.
The GitHub issue list is probably a good source of inspiration.
This could definitely be useful across models, which could be implemented simply by making the model an implicit filter.
See #109
Depends on the back-end implementation in