INDEX

Explanations

references to radio stations and shows

oai_token-act-pair · gpt-3.5-turbo

mentions of "Radio."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 advoc

-0.78

 virtuous

-0.75

 advant

-0.74

 strokes

-0.72

 cher

-0.69

 incorpor

-0.67

 vertex

-0.66

cus

-0.65

 attachments

-0.64

 Severus

-0.64

POSITIVE LOGITS

 Radio

3.89

Radio

3.15

radio

2.69

 radio

2.63

 radios

2.00

 Broadcasting

1.59

 Television

1.56

RAD

1.42

FM

1.37

 Audio

1.30

Activations Density 0.009%

references to radio stations and shows

mentions of "Radio."

No Comments

No Known Activations