INDEX

Explanations

phrases related to actions done by influential or authoritative figures

oai_token-act-pair · gpt-3.5-turbo

words related to ridicule or pranks

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

 Sacrament

-0.75

INESS

-0.70

 Winds

-0.69

croft

-0.65

born

-0.65

 Cheong

-0.64

 EntityItem

-0.62

 Civil

-0.61

 Americas

-0.59

 Raider

-0.59

POSITIVE LOGITS

imes

1.11

imate

1.08

uning

1.05

arians

1.05

icer

1.03

udence

1.01

atical

1.00

icks

0.99

acking

0.97

unn

0.97

Activations Density 0.004%

phrases related to actions done by influential or authoritative figures

words related to ridicule or pranks

No Comments

No Known Activations