INDEX

Explanations

the word "hugging" or variations of it

oai_token-act-pair · gpt-3.5-turbo

words that are reminiscent of the concept of "ugliness" or negative qualities

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 pole

-0.71

 forecast

-0.67

 realism

-0.67

hel

-0.66

 Planning

-0.64

CCP

-0.63

 Matthews

-0.61

 panel

-0.61

 forwards

-0.60

Sky

-0.60

POSITIVE LOGITS

ug

4.40

ugs

2.13

uga

1.89

uge

1.87

ugu

1.87

UG

1.87

ugi

1.75

ugal

1.71

ugen

1.67

ugg

1.41

Activations Density 0.010%

the word "hugging" or variations of it

words that are reminiscent of the concept of "ugliness" or negative qualities

No Comments

No Known Activations