INDEX

Explanations

sequences of characters "ice" with a high activation value

oai_token-act-pair · gpt-3.5-turbo

references to "ice" or words related to "ice" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

merce

-0.86

 scrut

-0.85

¥ŀ

-0.78

 referen

-0.74

sembly

-0.70

uilt

-0.68

semb

-0.68

 citiz

-0.67

cknow

-0.67

 newcom

-0.67

POSITIVE LOGITS

ice

1.40

lli

1.10

llular

0.98

llo

0.95

lla

0.94

ices

0.89

xp

0.79

 Cust

0.75

cone

0.75

ptive

0.72

Activations Density 0.008%

sequences of characters "ice" with a high activation value

references to "ice" or words related to "ice" in various contexts

No Comments

No Known Activations