INDEX

Explanations

names of specific individuals

oai_token-act-pair · gpt-3.5-turbo

keywords related to individuals in power and identity

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

•contextual features

completes words that start with a 'tam', 'hom', 'rel', 'lad', etc token

No Comments

Negative Logits

Cola

-0.72

 wiret

-0.67

 ACTIONS

-0.67

ĵĺ

-0.66

âķĲâķĲ

-0.64

 Alto

-0.64

>>>>

-0.62

 township

-0.61

 Windsor

-0.59

 succeeding

-0.58

POSITIVE LOGITS

andise

0.89

meric

0.75

ensional

0.74

iple

0.68

hematic

0.66

ificant

0.65

inarily

0.65

ovy

0.64

orage

0.64

achev

0.62

Activations Density 0.264%

names of specific individuals

keywords related to individuals in power and identity

No Comments

No Known Activations