INDEX

Explanations

conjunctions or words connected to cause and effect relationships

oai_token-act-pair · gpt-3.5-turbo

conjunctions that indicate causative or consequent relationships

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

estern

-0.72

lication

-0.71

 Bastard

-0.69

ãĥ¯ãĥ³

-0.66

 Latest

-0.62

ochet

-0.62

iren

-0.62

Own

-0.62

wic

-0.60

ATT

-0.60

POSITIVE LOGITS

 prevents

1.39

 reduces

1.30

 undermines

1.28

 enhances

1.27

 hinder

1.27

 lessen

1.23

 inhibits

1.22

 enables

1.21

 discourage

1.20

 exacerbate

1.19

Activations Density 0.279%

conjunctions or words connected to cause and effect relationships

conjunctions that indicate causative or consequent relationships

No Comments

No Known Activations