INDEX

Explanations

mentions of political figures and leaders

oai_token-act-pair · gpt-3.5-turbo

references to political figures and governmental positions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

Cth

-0.66

 Klan

-0.64

 Sioux

-0.64

 Appalachian

-0.64

NK

-0.63

philis

-0.63

opter

-0.62

 transsexual

-0.61

CVE

-0.61

 mole

-0.60

POSITIVE LOGITS

 chancellor

0.90

 cabinet

0.86

 Chancellor

0.83

 Finance

0.82

 Hollande

0.82

Chance

0.82

Prime

0.82

dinand

0.81

 ministers

0.80

 Premier

0.80

Activations Density 0.351%

mentions of political figures and leaders

references to political figures and governmental positions

No Comments

No Known Activations