INDEX

Explanations

negations or refusals

oai_token-act-pair · gpt-3.5-turbo

phrases emphasizing negation or the absence of something

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

urous

-0.64

essen

-0.60

 gaze

-0.59

tle

-0.59

aus

-0.58

irds

-0.57

eline

-0.57

heid

-0.57

 ambition

-0.56

ilde

-0.55

POSITIVE LOGITS

NOT

3.37

NOT

2.22

 NEVER

1.98

 ONLY

1.79

 ALWAYS

1.70

 ALSO

1.69

 WITHOUT

1.51

 THEN

1.50

 VERY

1.50

 REALLY

1.46

Activations Density 0.010%

negations or refusals

phrases emphasizing negation or the absence of something

No Comments

No Known Activations