INDEX

Explanations

phrases related to following instructions or moral behavior

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

elements related to NPR branding and publication information

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 12-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_post

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_post

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

 Mata

-0.78

ect

-0.75

 Guardiola

-0.72

 Silva

-0.68

 Expend

-0.68

Suk

-0.67

Tec

-0.66

 steroids

-0.65

 Sims

-0.64

ðŁĳ

-0.64

POSITIVE LOGITS

NPR

1.93

NPR

1.92

toggle

1.05

PBS

1.01

 listeners

0.93

 listener

0.86

Interview

0.85

Music

0.84

 Listen

0.84

 Transcript

0.83

Activations Density 0.086%

phrases related to following instructions or moral behavior

elements related to NPR branding and publication information

No Comments

No Known Activations