INDEX

Explanations

references to news outlets or news-related terms

oai_token-act-pair · gpt-3.5-turbo

instances of the word "news" and its variants

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 equival

-0.69

ãĤ©

-0.68

argon

-0.67

é¾

-0.66

verages

-0.63

DEBUG

-0.63

DoS

-0.63

IPM

-0.61

ashtra

-0.61

pmwiki

-0.58

POSITIVE LOGITS

leans

0.79

izons

0.74

angled

0.70

pires

0.70

pac

0.65

lisher

0.65

phrine

0.65

enment

0.64

ilings

0.64

enhagen

0.63

Activations Density 0.155%

references to news outlets or news-related terms

instances of the word "news" and its variants

No Comments

No Known Activations