INDEX

Explanations

references to malicious activities or intentions

oai_token-act-pair · gpt-3.5-turbo

references to malicious actions or intent, particularly in a cybersecurity context

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

ĸļ

-1.21

arist

-0.87

ills

-0.80

ļéĨĴ

-0.76

hetti

-0.74

ADRA

-0.70

akeru

-0.70

illy

-0.70

ucket

-0.69

alist

-0.69

POSITIVE LOGITS

ly

1.05

 intent

1.01

 payload

0.86

 malicious

0.86

 mischief

0.81

 behaviour

0.80

icious

0.79

vertising

0.78

 behavi

0.77

 activity

0.76

Activations Density 0.010%

references to malicious activities or intentions

references to malicious actions or intent, particularly in a cybersecurity context

No Comments

No Known Activations