INDEX

Explanations

the word "pit" or words related to pits

oai_token-act-pair · gpt-3.5-turbo

mentions of a specific location or venue

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

anwhile

-0.93

 Idle

-0.73

å¥³

-0.69

Ń·

-0.69

ittal

-0.67

inding

-0.63

erry

-0.63

demand

-0.62

eh

-0.62

ITED

-0.62

POSITIVE LOGITS

Pit

1.01

cher

0.97

 Slug

0.87

pit

0.87

crow

0.82

iless

0.81

Osw

0.80

ozo

0.79

chers

0.78

cair

0.78

Activations Density 0.004%

the word "pit" or words related to pits

mentions of a specific location or venue

No Comments

No Known Activations