INDEX

Explanations

phrases signaling permission, advice, or assistance.

oai_token-act-pair · gpt-3.5-turbo

phrases emphasizing the idea of not needing to do something or feeling obligated

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

Cyp

-0.75

Motion

-0.72

interstitial

-0.66

Eva

-0.66

iple

-0.65

wikipedia

-0.64

cember

-0.63

cade

-0.60

 Crate

-0.60

handled

-0.60

POSITIVE LOGITS

 worry

1.40

 bother

1.14

 anymore

1.07

 wait

0.97

 rely

0.93

 sacrifice

0.88

 fret

0.87

 stare

0.86

 suffer

0.86

 confront

0.84

Activations Density 0.067%

**phrases signaling permission, advice, or assistance.**

phrases emphasizing the idea of not needing to do something or feeling obligated

No Comments

No Known Activations

phrases signaling permission, advice, or assistance.