INDEX

Explanations

references to the word "Char" followed by a digit

oai_token-act-pair · gpt-3.5-turbo

mentions of a specific brand or character

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

MW

-0.87

 benchmark

-0.70

ively

-0.70

udo

-0.67

 prem

-0.67

OIL

-0.66

uckland

-0.65

Bench

-0.65

wave

-0.65

peak

-0.64

POSITIVE LOGITS

 Char

3.88

Char

2.79

char

1.66

 Charm

1.62

 char

1.61

 CHAR

1.52

 Chau

1.50

 Charity

1.39

Chr

1.28

Cha

1.22

Activations Density 0.011%

references to the word "Char" followed by a digit

mentions of a specific brand or character

No Comments

No Known Activations