INDEX

Explanations

phrases related to forums or conferences

oai_token-act-pair · gpt-3.5-turbo

mentions of various forums, particularly the World Economic Forum

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

 sing

-0.78

 paycheck

-0.69

 tear

-0.69

arted

-0.68

 apples

-0.65

carry

-0.64

 shakes

-0.62

miss

-0.62

externalActionCode

-0.61

ling

-0.61

POSITIVE LOGITS

 Forum

3.85

forum

2.00

 forum

1.91

 Forums

1.88

 forums

1.63

 Discussion

1.32

forums

1.22

 Dialogue

1.19

 Coliseum

1.14

 Workshop

1.09

Activations Density 0.006%

phrases related to forums or conferences

mentions of various forums, particularly the World Economic Forum

No Comments

No Known Activations