INDEX

Explanations

phrases related to evoking emotions or reactions

oai_token-act-pair · gpt-3.5-turbo

instances of the root "ev" suggesting avoidance or evading actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

WAY

-0.72

Ô

-0.69

WAYS

-0.69

 Emirates

-0.68

Ĥ¬

-0.63

 Retail

-0.62

 Franks

-0.62

 Responsibility

-0.61

SHARE

-0.61

Bei

-0.61

POSITIVE LOGITS

icted

1.20

asive

1.15

idences

1.06

isions

1.04

oked

1.04

ictions

1.02

itability

1.01

olve

1.00

ocative

0.99

idential

0.99

Activations Density 0.006%

phrases related to evoking emotions or reactions

instances of the root "ev" suggesting avoidance or evading actions

No Comments

No Known Activations