INDEX

Explanations

words related to intentional harm or deceitful actions

oai_token-act-pair · gpt-3.5-turbo

instances of the word "screw" and its variations

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

åī

-0.79

usable

-0.74

CTV

-0.71

vation

-0.69

CVE

-0.65

outh

-0.65

Interstitial

-0.64

obook

-0.64

 sovere

-0.63

icipated

-0.63

POSITIVE LOGITS

driver

1.31

drivers

1.11

hole

0.94

 Whedon

0.94

 screws

0.91

holes

0.88

ball

0.85

 screw

0.83

balls

0.82

nuts

0.76

Activations Density 0.014%

words related to intentional harm or deceitful actions

instances of the word "screw" and its variations

No Comments

No Known Activations