INDEX

Explanations

politically charged words and phrases; specifically, it seems to highlight strong or forceful statements

oai_token-act-pair · gpt-3.5-turbo

repeated characters or stylized characters in text

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

 disadvant

-0.86

 Gutenberg

-0.83

 mathemat

-0.81

 misunder

-0.70

inav

-0.69

geries

-0.69

merce

-0.68

 carbohyd

-0.67

 whiff

-0.64

raviolet

-0.63

POSITIVE LOGITS

ï¸ı

1.17

uth

0.95

å¥³

0.90

Â¯Â¯

0.88

ï¸

0.87

Â§

0.87

ution

0.81

ãģ®éŃĶ

0.81

ãĤĭ

0.80

Â·Â·

0.78

Activations Density 0.387%

politically charged words and phrases; specifically, it seems to highlight strong or forceful statements

repeated characters or stylized characters in text

No Comments

No Known Activations