INDEX

Explanations

text related to harmful, intentional actions

oai_token-act-pair · gpt-3.5-turbo

terms related to malicious intent or harmful actions

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

ĸļ

-1.26

arist

-0.86

ills

-0.81

akeru

-0.76

gdala

-0.73

ļéĨĴ

-0.72

illy

-0.71

hene

-0.71

Ã¨re

-0.69

blance

-0.69

POSITIVE LOGITS

ly

1.23

 intent

1.06

 mischief

0.87

 behaviour

0.87

 activity

0.84

 behavi

0.82

 payload

0.82

icious

0.82

LY

0.81

 malicious

0.80

Activations Density 0.021%

text related to harmful, intentional actions

terms related to malicious intent or harmful actions

No Comments

No Known Activations