INDEX

Explanations

the word "Hell"

oai_token-act-pair · gpt-3.5-turbo

mentions of "Hell" in various contexts

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

æĸ¹

-0.85

abet

-0.82

PsyNetMessage

-0.80

æ°

-0.79

æī

-0.78

å¤

-0.73

riad

-0.71

Marginal

-0.71

è¿

-0.71

éĹ

-0.71

POSITIVE LOGITS

enic

1.03

 Hell

0.94

ibur

0.92

hound

0.92

ocaust

0.86

cats

0.84

anger

0.84

wig

0.81

bender

0.80

bourne

0.80

Activations Density 0.011%

the word "Hell"

mentions of "Hell" in various contexts

No Comments

No Known Activations