INDEX

Explanations

references to former individuals in the political and public sphere

oai_token-act-pair · gpt-3.5-turbo

references to individuals with the title "Former."

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

antics

-0.90

otle

-0.86

acht

-0.85

anguage

-0.77

achus

-0.76

matic

-0.76

shapeshifter

-0.74

utterstock

-0.73

anche

-0.72

ammy

-0.72

POSITIVE LOGITS

 Yugoslav

0.90

 Yugoslavia

0.85

ly

0.80

 Vice

0.79

 President

0.78

 Deputy

0.76

bies

0.75

 Soviet

0.74

foundland

0.74

 vice

0.73

Activations Density 0.034%

references to former individuals in the political and public sphere

references to individuals with the title "Former."

No Comments

No Known Activations