INDEX

Explanations

mentions of actions related to saving or protection

oai_token-act-pair · gpt-3.5-turbo

instances of the word "save" and its variations related to rescue or preservation

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

Intern

-0.67

yond

-0.66

kt

-0.66

VP

-0.63

quartered

-0.63

FG

-0.61

pora

-0.60

 sclerosis

-0.59

Rush

-0.58

KER

-0.58

POSITIVE LOGITS

Save

0.93

 Save

0.79

 Saving

0.77

saving

0.75

 Sanctuary

0.74

 Lives

0.74

anza

0.73

luc

0.70

usc

0.69

 lives

0.66

Activations Density 0.026%

mentions of actions related to saving or protection

instances of the word "save" and its variations related to rescue or preservation

No Comments

No Known Activations