INDEX

Explanations

words related to power, control, and authorization

oai_token-act-pair · gpt-3.5-turbo

references to the concept of authority

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

 Lovely

-0.79

 Resist

-0.75

 Taste

-0.75

 Leica

-0.72

 Sail

-0.70

Fir

-0.70

Il

-0.68

WM

-0.68

 Shal

-0.68

 Rift

-0.66

POSITIVE LOGITS

 authority

1.38

Reviewer

1.15

 confir

0.99

authorized

0.93

 estab

0.91

accompan

0.88

 authorities

0.87

 delegated

0.85

abulary

0.85

 vested

0.85

Activations Density 0.011%

words related to power, control, and authorization

references to the concept of authority

No Comments

No Known Activations