INDEX

Explanations

references to positions or roles

oai_token-act-pair · gpt-3.5-turbo

references to a specific "position" 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 @ 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

ombies

-0.86

 bells

-0.70

angel

-0.64

 Anonymous

-0.64

jah

-0.64

 THANK

-0.64

NEWS

-0.63

hall

-0.63

icious

-0.61

bey

-0.61

POSITIVE LOGITS

 position

3.83

 positions

2.87

 Position

2.67

Position

2.21

position

2.01

 stance

1.83

 positioning

1.78

 posture

1.62

Pos

1.44

 role

1.39

Activations Density 0.014%

references to positions or roles

references to a specific "position" in various contexts

No Comments

No Known Activations