INDEX

Explanations

phrases related to the concept of seizing control or physical contact

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "hold."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Audit

-0.77

TeX

-0.70

endez

-0.66

 Airways

-0.63

xon

-0.63

ople

-0.62

acha

-0.62

 Chou

-0.62

icycle

-0.62

urses

-0.62

POSITIVE LOGITS

hold

1.42

holder

0.99

holding

0.99

backs

0.97

ings

0.96

holders

0.94

back

0.93

erness

0.87

fell

0.83

fast

0.82

Activations Density 0.027%

phrases related to the concept of seizing control or physical contact

occurrences of the word "hold."

No Comments

No Known Activations