INDEX

Explanations

words related to location names or entities

oai_token-act-pair · gpt-3.5-turbo

suffixes that denote actions or processes

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 disbel

-0.64

 Salem

-0.61

 awake

-0.59

soType

-0.58

 clinics

-0.56

PRES

-0.56

 clinic

-0.55

 metast

-0.55

 depress

-0.55

 Mens

-0.54

POSITIVE LOGITS

ion

4.56

ions

2.89

ION

2.55

ioned

2.36

ional

2.12

ione

1.90

IONS

1.85

ionic

1.66

iona

1.48

iod

1.46

Activations Density 0.012%

words related to location names or entities

suffixes that denote actions or processes

No Comments

No Known Activations