INDEX

Explanations

names, particularly "Dodd" and "Newark"

oai_token-act-pair · gpt-3.5-turbo

references to specific names and entities, particularly related to individuals and locations

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

 Porn

-0.74

pel

-0.73

Vict

-0.71

 porn

-0.70

 favourites

-0.69

 pellets

-0.69

itas

-0.68

 Hungry

-0.68

 Nuggets

-0.68

 aval

-0.68

POSITIVE LOGITS

 Dodd

2.62

itary

1.47

 Byrne

1.31

 Newport

1.27

 Newark

1.21

 premature

1.20

 Hond

1.20

ohn

1.11

erred

1.10

hap

1.05

Activations Density 0.045%

names, particularly "Dodd" and "Newark"

references to specific names and entities, particularly related to individuals and locations

No Comments

No Known Activations