INDEX

Explanations

words related to terrorism and counter-terrorism

oai_token-act-pair · gpt-3.5-turbo

references to terrorism

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 acknow

-0.77

saf

-0.76

BYU

-0.76

aples

-0.74

actionDate

-0.73

 circumcised

-0.71

Davis

-0.70

 handwriting

-0.70

ruby

-0.69

 divorced

-0.69

POSITIVE LOGITS

 Terror

2.96

Terror

2.12

terror

1.33

 Troll

1.22

 Horror

1.11

 Uzbek

1.04

 Terrorism

1.04

 terror

1.02

oreal

1.02

 Cann

1.00

Activations Density 0.052%

words related to terrorism and counter-terrorism

references to terrorism

No Comments

No Known Activations