INDEX

Explanations

adjectives with positive or negative connotations, with a preference for negative ones

oai_token-act-pair · gpt-3.5-turbo

words and phrases related to truth, reasoning, and contradictions

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

gdala

-0.84

 Flavoring

-0.71

 throats

-0.71

 virginity

-0.63

 contacts

-0.60

 purse

-0.58

deen

-0.58

 packs

-0.58

 slot

-0.58

 nickname

-0.58

POSITIVE LOGITS

actly

1.12

ivable

1.12

urable

1.05

icable

1.03

ceivable

1.02

inently

1.02

vable

1.01

gru

1.01

itatively

1.00

izable

1.00

Activations Density 0.240%

adjectives with positive or negative connotations, with a preference for negative ones

words and phrases related to truth, reasoning, and contradictions

No Comments

No Known Activations