INDEX

Explanations

words related to death or killing

oai_token-act-pair · gpt-3.5-turbo

instances of people being killed

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

wcsstore

-0.81

arity

-0.80

issance

-0.70

issue

-0.68

worldly

-0.67

 Plex

-0.66

 Collider

-0.65

rium

-0.64

yrinth

-0.64

arist

-0.61

POSITIVE LOGITS

 spree

0.92

 slain

0.75

locked

0.74

 bystand

0.72

nsics

0.72

switch

0.72

 Massacre

0.72

 rampage

0.70

linger

0.70

 bystanders

0.69

Activations Density 0.031%

words related to death or killing

instances of people being killed

No Comments

No Known Activations