INDEX

Explanations

code execution context

The neuron is picking out “code‐style” tokens—things like variable names (including identifiers with underscores) and literal numbers in code snippets.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 Việc

0.70

 事業

0.68

 onRequest

0.66

 정치

0.64

 Políticas

0.64

 プロ

0.64

Já

0.64

 Riot

0.64

 ビジネス

0.64

 사업

0.63

POSITIVE LOGITS

_,

0.81

idx

0.77

_=

0.74

imgs

0.71

tmp

0.71

_,

0.71

 data

0.69

data

0.68

npy

0.68

 rmse

0.67

Activations Density 0.000%