INDEX

Explanations

words related to firearms and violent actions, such as 'fired', 'shooting', and 'gun'

oai_token-act-pair · gpt-3.5-turbo

actions related to shooting or discharging firearms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

Label

-0.74

akuya

-0.72

Climate

-0.71

Redditor

-0.67

algia

-0.67

Score

-0.67

fty

-0.64

DonaldTrump

-0.64

 Obesity

-0.63

Pont

-0.63

POSITIVE LOGITS

 gunshots

1.00

balls

0.99

 gunfire

0.98

 volley

0.95

 bullets

0.94

 projectiles

0.93

bomb

0.93

 cannons

0.93

 missiles

0.90

 ammunition

0.90

Activations Density 0.053%

words related to firearms and violent actions, such as 'fired', 'shooting', and 'gun'

actions related to shooting or discharging firearms

No Comments

No Known Activations