INDEX

Explanations

instances where something has been formerly known as another term

oai_token-act-pair · gpt-3.5-turbo

references to entities or concepts that were previously known by a different name or identity

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

anche

-0.77

antics

-0.77

pled

-0.72

atic

-0.69

jriwal

-0.68

ences

-0.67

eer

-0.66

sure

-0.65

ilation

-0.65

atics

-0.64

POSITIVE LOGITS

 formerly

0.95

 obsolete

0.92

 defunct

0.89

 vacated

0.82

 styled

0.82

 incarcerated

0.79

etheless

0.79

 extinct

0.78

 dubbed

0.78

 redes

0.78

Activations Density 0.003%

instances where something has been formerly known as another term

references to entities or concepts that were previously known by a different name or identity

No Comments

No Known Activations