INDEX

Explanations

phrases related to conflict, attack, and controversial actions

oai_token-act-pair · gpt-3.5-turbo

phrases involving aggressive or coercive actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Lets

-0.73

ETS

-0.70

 Going

-0.67

 Came

-0.66

Tid

-0.64

Peb

-0.64

gettable

-0.64

gets

-0.63

 Spears

-0.63

 Rings

-0.63

POSITIVE LOGITS

 degrade

1.12

rogen

1.08

 humili

1.07

 intimidate

1.05

igmat

1.04

rogens

0.98

 distort

0.97

 harass

0.96

 punish

0.96

 discredit

0.95

Activations Density 0.120%

phrases related to conflict, attack, and controversial actions

phrases involving aggressive or coercive actions

No Comments

No Known Activations