INDEX

Explanations

mentions of popular culture references such as 'James Bond' and interactive elements like questions

oai_token-act-pair · gpt-3.5-turbo

phrases that suggest a playful or casual tone

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.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

 analogous

-0.72

 overlap

-0.69

 UNCLASSIFIED

-0.69

 fluct

-0.69

Newsletter

-0.68

 advers

-0.67

 generalized

-0.67

 emphasizes

-0.67

etc

-0.67

 overarching

-0.66

POSITIVE LOGITS

TOR

0.78

 rejoice

0.76

Redditor

0.74

 Pinball

0.74

 Morty

0.73

 Adelaide

0.73

CTV

0.70

 TODAY

0.69

dies

0.69

 Mehran

0.67

Activations Density 0.811%

mentions of popular culture references such as 'James Bond' and interactive elements like questions

phrases that suggest a playful or casual tone

No Comments

No Known Activations