INDEX

Explanations

references to violent crimes, specifically murder

oai_token-act-pair · gpt-3.5-turbo

references to murder and related terminology

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

orem

-0.84

SEN

-0.71

æ©Ł

-0.68

SY

-0.67

malink

-0.66

¶ħ

-0.65

ukong

-0.65

sonian

-0.64

NOT

-0.64

ãĥķãĤ©

-0.64

POSITIVE LOGITS

abilia

0.85

ocaust

0.84

itual

0.80

 vigil

0.77

apego

0.76

 prosec

0.74

rarily

0.74

 Chains

0.74

esville

0.74

 Corpse

0.73

Activations Density 0.112%

references to violent crimes, specifically murder

references to murder and related terminology

No Comments

No Known Activations