INDEX

Explanations

references to the concept of a "shield"

oai_token-act-pair · gpt-3.5-turbo

references to protective or defensive mechanisms

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

 Vide

-0.78

ETA

-0.77

uni

-0.75

ovie

-0.75

pheus

-0.72

PM

-0.71

 Helpful

-0.71

orie

-0.70

gres

-0.70

orig

-0.69

POSITIVE LOGITS

 shields

1.24

 shield

1.22

 shielding

0.97

Shield

0.97

 Shield

0.93

maid

0.90

shield

0.90

 defences

0.85

 curtain

0.83

 Shields

0.81

Activations Density 0.005%

references to the concept of a "shield"

references to protective or defensive mechanisms

No Comments

No Known Activations