INDEX

Explanations

phrases containing names of people in positions of authority

oai_token-act-pair · gpt-3.5-turbo

references to individuals in positions of authority or intelligence roles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

opian

-0.75

 Jagu

-0.71

[|

-0.70

 nostalg

-0.69

egal

-0.69

 Twins

-0.68

nings

-0.65

ahime

-0.65

VALUE

-0.64

winner

-0.63

POSITIVE LOGITS

 Mattis

0.84

 Comey

0.79

 Flynn

0.78

 Merrill

0.76

 Gates

0.73

iden

0.73

ATF

0.72

 Holder

0.72

 Tillerson

0.72

 DeVos

0.71

Activations Density 0.129%

phrases containing names of people in positions of authority

references to individuals in positions of authority or intelligence roles

No Comments

No Known Activations