INDEX

Explanations

telescope, bleeding, visual fields

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

rost

-0.09

 Pont

-0.09

Compat

-0.09

 surgical

-0.09

CUT

-0.08

 Latter

-0.08

 Boys

-0.08

 AXIS

-0.08

 Surgical

-0.08

 outr

-0.08

POSITIVE LOGITS

ILD

0.10

 buzzing

0.10

[*

0.10

NPC

0.09

 HARD

0.09

 gains

0.09

 experimenting

0.09

 Arrow

0.09

 dere

0.09

 plateau

0.09

Activations Density 0.130%