INDEX

Explanations

references to past connections or relationships with someone or something

oai_token-act-pair · gpt-3.5-turbo

references to previous roles or statuses of individuals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

achus

-0.98

otle

-0.91

anguage

-0.83

andise

-0.81

onso

-0.76

antics

-0.75

allery

-0.74

ourced

-0.73

ghai

-0.73

acht

-0.72

POSITIVE LOGITS

 Yugoslavia

0.96

 Yugoslav

0.95

 Soviet

0.82

 captives

0.78

 comrade

0.76

 dictator

0.76

Lum

0.75

 President

0.75

 wartime

0.75

 comrades

0.75

Activations Density 0.029%

references to past connections or relationships with someone or something

references to previous roles or statuses of individuals

No Comments

No Known Activations