INDEX

Explanations

possessive pronoun followed by noun

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

 intent

-0.10

inger

-0.09

Pry

-0.09

lez

-0.09

 probe

-0.09

intent

-0.09

 Beaver

-0.08

 premise

-0.08

Fet

-0.08

INGER

-0.08

POSITIVE LOGITS

 responds

0.19

 respond

0.18

 responses

0.17

 responded

0.17

 responding

0.17

 reacts

0.16

 reaction

0.16

 reacted

0.16

 reactions

0.16

 response

0.15

Activations Density 0.004%