INDEX

Explanations

phrases related to transitions in time or stages

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 Wikileaks

-0.78

 confession

-0.70

 cryptography

-0.66

 Reddit

-0.66

 Coinbase

-0.65

 IMAGES

-0.64

 informant

-0.63

Wiki

-0.63

 informants

-0.63

 Allaah

-0.62

POSITIVE LOGITS

Finish

0.68

ãĤ©

0.66

 operate

0.66

ãĥīãĥ©

0.65

 overrun

0.63

 exting

0.62

ighty

0.61

 reintrodu

0.60

ageddon

0.60

 wiped

0.60

Activations Density 0.810%

phrases related to transitions in time or stages

No Comments

No Known Activations