INDEX

Explanations

instances where someone is being directly addressed or summoned

oai_token-act-pair · gpt-3.5-turbo

instances of the word "called."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.9.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

olitics

-0.82

edia

-0.82

iland

-0.77

bilt

-0.76

feat

-0.76

enture

-0.71

yip

-0.70

edom

-0.66

ebin

-0.65

isphere

-0.64

POSITIVE LOGITS

 upon

0.96

 forth

0.83

 Attention

0.72

 attention

0.71

 into

0.70

 bluff

0.67

 responsible

0.66

oused

0.66

ingen

0.64

up

0.63

Activations Density 0.054%

instances where someone is being directly addressed or summoned

instances of the word "called."

No Comments

No Known Activations