INDEX

Explanations

phrases related to locations, especially cities and states

oai_token-act-pair · gpt-3.5-turbo

references to the word "Worth" and associated locations or contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 intervene

-0.69

enc

-0.67

 Prime

-0.67

ubi

-0.65

ins

-0.64

sub

-0.64

 Bolshe

-0.63

 chees

-0.63

mud

-0.62

ice

-0.60

POSITIVE LOGITS

 Worth

3.03

 Lauderdale

1.93

worth

1.36

patrick

1.25

 Haven

1.20

 McGregor

1.10

ardy

1.05

 Myers

1.05

ayson

1.03

 Grayson

1.02

Activations Density 0.021%

phrases related to locations, especially cities and states

references to the word "Worth" and associated locations or contexts

No Comments

No Known Activations