INDEX

Explanations

words related to something that is obvious, intentional, and usually morally wrong or harmful

oai_token-act-pair · gpt-3.5-turbo

instances of the word "blatant" and its variations, indicating clear and obvious issues or hypocrisy

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

nesota

-0.92

ulton

-0.91

arios

-0.85

otos

-0.79

erves

-0.79

oan

-0.77

arters

-0.76

agine

-0.75

hner

-0.73

ould

-0.72

POSITIVE LOGITS

 blatant

0.91

 disregard

0.85

 blatantly

0.82

 brazen

0.79

iary

0.79

 shameless

0.78

 misrepresent

0.78

 viol

0.76

 contradictions

0.75

ãĥĨ

0.74

Activations Density 0.011%

words related to something that is obvious, intentional, and usually morally wrong or harmful

instances of the word "blatant" and its variations, indicating clear and obvious issues or hypocrisy

No Comments

No Known Activations