INDEX

Explanations

words or prefixes related to evasion or avoidance

oai_token-act-pair · gpt-3.5-turbo

instances of the prefix "ev," indicating various actions or phenomena associated with evasion or avoidance

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

-0.74

WAYS

-0.68

Ô

-0.66

 Emirates

-0.65

cone

-0.65

 Responsibility

-0.64

matically

-0.63

 Franks

-0.61

 Principle

-0.60

WARE

-0.60

POSITIVE LOGITS

icted

1.24

asive

1.17

olve

1.09

isions

1.08

ictions

1.08

idences

1.05

idently

1.03

itability

1.03

ocative

1.03

oked

1.02

Activations Density 0.008%

words or prefixes related to evasion or avoidance

instances of the prefix "ev," indicating various actions or phenomena associated with evasion or avoidance

No Comments

No Known Activations