INDEX

Explanations

recurring words and phrases that indicate ongoing engagement

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Configuration

ctigges/pythia-70m-deduped__res-sm_processed/2-res-sm

Prompts (Dashboard)

32,768 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

32,768

Data Type

torch.float32

Hook Name

blocks.2.hook_resid_post

Hook Layer

Architecture

standard

Context Size

128

Dataset

EleutherAI/the_pile_deduplicated

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

arte

-1.59

![**

-1.44

encing

-1.44

ruck

-1.41

ester

-1.40

ault

-1.40

arrest

-1.39

 Spearman

-1.37

entes

-1.36

ator

-1.35

POSITIVE LOGITS

¿½

2.56

ª

2.24

Ľ

2.21

Īĺ

2.17

³

2.15

±

2.15

Ļ

2.14

¤

2.14

ĥ½

2.02

ī

1.96

Activations Density 0.986%

recurring words and phrases that indicate ongoing engagement

No Comments

No Known Activations