INDEX

Explanations

years mentioned between 2001 and 2006

oai_token-act-pair · gpt-3.5-turbo

references to the year 2003

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

place

-0.70

ymes

-0.67

bow

-0.67

stakes

-0.66

pherd

-0.64

anches

-0.64

agues

-0.63

powers

-0.62

hop

-0.61

hma

-0.61

POSITIVE LOGITS

-'

0.83

ĸļ

0.81

å¹

0.79

Fuk

0.69

ãĥĥãĥī

0.69

iens

0.68

 onwards

0.63

ãĥ³ãĤ¸

0.63

rad

0.61

ritional

0.61

Activations Density 0.020%

years mentioned between 2001 and 2006

references to the year 2003

No Comments

No Known Activations