INDEX

Explanations

phrases related to political figures or locations, particularly "Downing Street."

oai_token-act-pair · gpt-3.5-turbo

references to specific locations and political figures

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

ropri

-0.73

rieg

-0.69

afia

-0.68

 Gutenberg

-0.67

netflix

-0.66

ilk

-0.65

ENDED

-0.65

olina

-0.64

eport

-0.63

asters

-0.63

POSITIVE LOGITS

 Downing

0.96

nuts

0.95

nut

0.87

stall

0.83

vous

0.82

tree

0.81

hurst

0.81

stein

0.80

 Britann

0.68

berg

0.67

Activations Density 0.026%

phrases related to political figures or locations, particularly "Downing Street."

references to specific locations and political figures

No Comments

No Known Activations