INDEX

Explanations

words related to names and titles, specifically people, organizations, and events

oai_token-act-pair · gpt-3.5-turbo

mentions of individuals or organizations, particularly those represented by certain titles or names

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

thumbnails

-0.62

 Hutch

-0.61

 Wonderland

-0.60

 Amend

-0.60

 Monroe

-0.60

 "$:/

-0.57

romeda

-0.56

 Hamm

-0.55

 Matter

-0.55

 Fraz

-0.55

POSITIVE LOGITS

ciples

1.18

cipled

1.16

cess

0.87

bably

0.85

ety

0.78

eton

0.76

incial

0.74

fecture

0.70

ciation

0.69

ongyang

0.69

Activations Density 0.064%

words related to names and titles, specifically people, organizations, and events

mentions of individuals or organizations, particularly those represented by certain titles or names

No Comments

No Known Activations