INDEX

Explanations

phrases related to a specific person, Joe Scarborough

oai_token-act-pair · gpt-3.5-turbo

references to the individual Joe Scarborough

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

phal

-0.92

istry

-0.78

endar

-0.76

ivated

-0.74

Ħ¢

-0.74

ment

-0.74

ĺ

-0.74

ph

-0.73

oci

-0.73

erie

-0.73

POSITIVE LOGITS

 Peaks

0.76

Sho

0.74

 Parenthood

0.70

 Sparkle

0.67

 Lakes

0.66

issance

0.66

Series

0.65

omething

0.63

 Burnett

0.63

éļ

0.62

Activations Density 0.052%

phrases related to a specific person, Joe Scarborough

references to the individual Joe Scarborough

No Comments

No Known Activations