INDEX

Explanations

phrases related to significant negative impacts or threats

oai_token-act-pair · gpt-3.5-turbo

terms associated with negative impacts or threats

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Creed

-0.60

ypes

-0.60

tein

-0.60

construct

-0.59

ynt

-0.58

 scaff

-0.57

ynthesis

-0.56

 choir

-0.55

hent

-0.54

 Conquer

-0.53

POSITIVE LOGITS

to

0.90

for

0.89

 unto

0.83

 thereto

0.80

 indeed

0.75

Interstitial

0.73

 financially

0.72

 downstream

0.71

 economically

0.71

 domestically

0.69

Activations Density 0.262%

phrases related to significant negative impacts or threats

terms associated with negative impacts or threats

No Comments

No Known Activations