INDEX

Explanations

mentions of being elected or re-elected into a position of authority

oai_token-act-pair · gpt-3.5-turbo

terms related to elections and electoral processes

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 condition

-0.69

 crate

-0.67

 fragment

-0.67

 Origin

-0.65

 Monster

-0.65

 pile

-0.64

 litter

-0.64

 field

-0.63

ness

-0.63

 subs

-0.63

POSITIVE LOGITS

elected

4.08

 elected

1.92

election

1.55

elect

1.35

appointed

1.29

 electing

1.26

educated

1.18

married

1.17

democratic

1.14

 democratically

1.14

Activations Density 0.017%

mentions of being elected or re-elected into a position of authority

terms related to elections and electoral processes

No Comments

No Known Activations