INDEX

Explanations

phrases that involve the concept of "former."

oai_token-act-pair · gpt-3.5-turbo

references to past states or conditions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

otle

-0.86

opped

-0.74

anders

-0.72

arity

-0.71

ammy

-0.70

oked

-0.70

achus

-0.67

ï¸ı

-0.66

aris

-0.65

Bott

-0.65

POSITIVE LOGITS

 Yugoslavia

1.18

 Yugoslav

1.06

 Soviet

0.98

 glory

0.93

 USSR

0.87

 slaves

0.83

 captives

0.81

 fashioned

0.79

 colonies

0.78

Soviet

0.76

Activations Density 0.034%

phrases that involve the concept of "former."

references to past states or conditions

No Comments

No Known Activations