Tag
This paper introduces a biologically motivated neural architecture that respects Dale's constraint—neurons and synapses have fixed sign—while still supporting backpropagation-like learning using non-negative error signals and local Hebbian updates, showing gains on Tiny ImageNet.
This paper investigates neural circuits in a sparse 8-layer Python transformer, finding dedicated circuitry for 106 programming concepts and decomposing them into concept-specific and token-driven components, with implications for understanding structural encoding in code models.
This paper investigates the internal mechanisms of LLM-as-a-judge, finding a shared Latent Evaluator sub-graph in mid-to-late MLPs across models that handles abstract judging, while format-specific terminal branches map the judgment to output tokens, revealing the cause of format-induced inconsistency.