INDEX

Explanations

the word "that" paired with a high activation value

oai_token-act-pair · gpt-3.5-turbo

the word "that" 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

 antioxid

-0.76

CLASSIFIED

-0.73

 newcom

-0.67

Interstitial

-0.62

ï¸

-0.62

hoff

-0.61

displayText

-0.60

é»Ĵ

-0.59

VIDEO

-0.59

OND

-0.58

POSITIVE LOGITS

 that

1.96

that

1.61

 THAT

1.45

 That

1.17

That

1.11

 thats

0.93

 which

0.83

 those

0.83

 this

0.81

 whose

0.75

Activations Density 0.735%

the word "that" paired with a high activation value

the word "that" in various contexts

No Comments

No Known Activations