INDEX

Explanations

references to someone taking an action or making a request towards another person

oai_token-act-pair · gpt-3.5-turbo

references to personal pronouns and their relationships to actions or requests

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 dawn

-0.63

 Course

-0.62

 Firm

-0.61

Nah

-0.59

AMI

-0.59

 Sina

-0.59

 Seller

-0.59

 Tale

-0.58

quickShipAvailable

-0.57

 Mandarin

-0.56

POSITIVE LOGITS

 hooked

0.90

 acquainted

0.77

ghan

0.75

nuts

0.74

agy

0.72

 fooled

0.71

 addicted

0.70

 nuts

0.70

ander

0.69

 deported

0.68

Activations Density 0.131%

references to someone taking an action or making a request towards another person

references to personal pronouns and their relationships to actions or requests

No Comments

No Known Activations