INDEX

Explanations

Roman numerals

oai_token-act-pair · gpt-3.5-turbo

occurrences of specific character patterns or symbols

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 manif

-0.87

 disadvant

-0.87

 misunder

-0.86

 levers

-0.83

 Vaugh

-0.80

 vulner

-0.78

geries

-0.76

 promoters

-0.75

 incorpor

-0.74

 fronts

-0.74

POSITIVE LOGITS

ï¸ı

1.38

 âĢº

0.92

cffffcc

0.90

âĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢâĶĢ

0.86

âĸ

0.85

×©

0.83

âĸ¬âĸ¬

0.83

HUD

0.82

×

0.80

STAR

0.80

Activations Density 0.106%

Roman numerals

occurrences of specific character patterns or symbols

No Comments

No Known Activations