INDEX

Explanations

phrases related to expressing contradiction or disagreement

oai_token-act-pair · gpt-3.5-turbo

negations

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

 around

-0.68

pal

-0.68

rig

-0.66

 shifts

-0.65

 troop

-0.63

 wagon

-0.63

 specialist

-0.60

iday

-0.60

 himself

-0.59

Cu

-0.58

POSITIVE LOGITS

Not

3.16

not

1.93

Not

1.88

NOT

1.79

Never

1.63

Nothing

1.57

Probably

1.53

Already

1.53

Only

1.50

Except

1.49

Activations Density 0.012%

phrases related to expressing contradiction or disagreement

negations

No Comments

No Known Activations