INDEX

Explanations

phrases indicating violent actions against individuals

oai_token-act-pair · gpt-3.5-turbo

instances of the word "killed" and related terms indicating death

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

rium

-0.77

Cola

-0.68

DragonMagazine

-0.68

idth

-0.63

ertain

-0.63

ĨĴ

-0.61

users

-0.61

ZX

-0.60

 CLSID

-0.60

 Libre

-0.60

POSITIVE LOGITS

by

0.84

 spree

0.82

 tragically

0.75

 rampage

0.71

 slain

0.71

heading

0.70

 gunned

0.70

nikov

0.69

abouts

0.68

 fighting

0.68

Activations Density 0.078%

phrases indicating violent actions against individuals

instances of the word "killed" and related terms indicating death

No Comments

No Known Activations