INDEX

Explanations

consistent patterns in usernames or profile information

oai_token-act-pair · gpt-3.5-turbo

numerical or rating information related to profiles or listings

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

 convictions

-0.81

 exposures

-0.79

 mortgages

-0.74

 loans

-0.73

 Opposition

-0.72

 disclosures

-0.68

 resolves

-0.68

ÂłÂł

-0.67

 leases

-0.67

 deposits

-0.65

POSITIVE LOGITS

mega

0.96

Anonymous

0.95

0.94

sam

0.89

dj

0.88

kay

0.88

eric

0.88

Luke

0.83

jad

0.83

jay

0.83

Activations Density 0.134%

consistent patterns in usernames or profile information

numerical or rating information related to profiles or listings

No Comments

No Known Activations