INDEX

Explanations

actions related to legal or medical outcomes with potential negative consequences

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

idth

-0.79

reci

-0.61

lished

-0.60

overty

-0.59

hov

-0.58

heny

-0.56

rouse

-0.56

icipated

-0.55

 RPGs

-0.55

 Peaks

-0.54

POSITIVE LOGITS

 anyway

1.00

 afterwards

0.98

 afterward

0.96

 anyways

0.95

 instantly

0.93

 promptly

0.92

 accordingly

0.92

 unanimously

0.91

 luckily

0.89

 shortly

0.88

Activations Density 0.464%

actions related to legal or medical outcomes with potential negative consequences

No Comments

No Known Activations