INDEX

Explanations

adjectives ending in '-tered'

oai_token-act-pair · gpt-3.5-turbo

words related to various forms of being "unfiltered" or "unsettled."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Instruments

-0.61

 question

-0.60

guard

-0.60

Ru

-0.57

 halftime

-0.56

 needle

-0.56

 Guard

-0.56

 Shuttle

-0.55

 staging

-0.55

 Releases

-0.55

POSITIVE LOGITS

tered

4.79

tering

2.97

ters

2.40

ter

1.75

ted

1.49

TER

1.47

terson

1.37

terness

1.30

gered

1.28

ched

1.26

Activations Density 0.013%

adjectives ending in '-tered'

words related to various forms of being "unfiltered" or "unsettled."

No Comments

No Known Activations