INDEX

Explanations

mentions of people being killed

oai_token-act-pair · gpt-3.5-turbo

instances of the word "killed" indicating death-related events

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

arist

-0.71

issue

-0.71

 Collider

-0.71

arity

-0.70

orers

-0.69

por

-0.69

endum

-0.66

rium

-0.66

 Plex

-0.65

Scot

-0.65

POSITIVE LOGITS

 spree

0.90

 killing

0.81

icides

0.80

 killings

0.78

 slain

0.75

killed

0.74

 killed

0.74

 rampage

0.74

 deaths

0.72

 Slaughter

0.71

Activations Density 0.030%

mentions of people being killed

instances of the word "killed" indicating death-related events

No Comments

No Known Activations