INDEX

Explanations

mentions of the word "soap" at various strengths of activation

oai_token-act-pair · gpt-3.5-turbo

references to soap operas

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

trl

-0.74

etermined

-0.69

umbered

-0.68

orders

-0.67

 landfall

-0.62

VEN

-0.61

 sworn

-0.61

uding

-0.60

CLASSIFIED

-0.60

 omission

-0.60

POSITIVE LOGITS

 opera

1.17

 oper

1.02

 soap

1.02

eless

0.87

bowl

0.81

 Leone

0.78

herer

0.77

utical

0.77

onica

0.77

pering

0.75

Activations Density 0.005%

mentions of the word "soap" at various strengths of activation

references to soap operas

No Comments

No Known Activations