American history tutors

The neuron fires strongly on tokens that look like technical identifiers—strings mixing letters, digits, punctuation (e.g. filenames, version numbers, domain names, constant names)—rather than natural-language words.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

halts

-1.23

fokus

-1.17

 capot

-1.14

quelle

-1.12

 rendement

-1.11

㎎

-1.09

bici

-1.09

 pistolet

-1.09

 douane

-1.09

akteur

-1.09

POSITIVE LOGITS

1.27

1.15

1.13

1.10

1.05

1.04

 manifestó

1.02

1.01

因为它

1.00

Activations Density 0.036%