INDEX

Explanations

dates or years

oai_token-act-pair · gpt-3.5-turbo

references to specific years

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

 contag

-0.70

 cumbers

-0.70

atility

-0.68

 showc

-0.67

ktop

-0.66

 encl

-0.66

creen

-0.66

 textual

-0.65

 confir

-0.65

 externalToEVAOnly

-0.64

POSITIVE LOGITS

long

1.09

ning

1.00

ns

0.96

ago

0.96

Ago

0.95

lings

0.94

 olds

0.91

book

0.89

nings

0.88

lies

0.82

Activations Density 0.103%

dates or years

references to specific years

No Comments

No Known Activations