INDEX

Explanations

words related to confidentiality and secrecy

oai_token-act-pair · gpt-3.5-turbo

references to sensitive information or topics

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

FIN

-0.83

LOAD

-0.75

 Wolver

-0.74

UNCH

-0.73

AUT

-0.72

 Truck

-0.72

INST

-0.70

aneers

-0.69

amaz

-0.68

swick

-0.68

POSITIVE LOGITS

 sensitive

1.17

ivities

1.07

mble

0.96

 sensit

0.80

ively

0.79

 sensitivity

0.77

itives

0.76

sensitive

0.76

ensitive

0.75

itiz

0.75

Activations Density 0.016%

words related to confidentiality and secrecy

references to sensitive information or topics

No Comments

No Known Activations