INDEX

Explanations

references to violent or harmful actions, often involving individuals or entities labeled as "killers"

oai_token-act-pair · gpt-3.5-turbo

references to "killer" in various contexts, especially relating to crime or wildlife

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

urn

-0.78

ional

-0.78

bles

-0.77

ational

-0.74

arb

-0.71

auri

-0.71

rity

-0.70

REF

-0.69

adr

-0.69

ourced

-0.69

POSITIVE LOGITS

 killer

1.32

 killers

1.13

killer

1.09

 Killer

1.07

 whales

0.84

knife

0.76

 whale

0.75

killers

0.75

 assassin

0.75

ipop

0.73

Activations Density 0.009%

references to violent or harmful actions, often involving individuals or entities labeled as "killers"

references to "killer" in various contexts, especially relating to crime or wildlife

No Comments

No Known Activations