INDEX

Explanations

phrases related to measurements or technical specifications

oai_token-act-pair · gpt-3.5-turbo

references to individuals and specific names

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

 Reform

-0.78

 Ezra

-0.64

 recycle

-0.63

ASH

-0.62

 Andromeda

-0.60

obin

-0.60

 Standard

-0.60

 Letter

-0.59

 recycling

-0.59

 obscure

-0.58

POSITIVE LOGITS

vic

4.24

cu

1.91

voy

1.46

const

1.28

vic

1.18

idon

1.15

wic

1.14

Pont

1.12

vag

1.10

anic

1.09

Activations Density 0.026%

phrases related to measurements or technical specifications

references to individuals and specific names

No Comments

No Known Activations