INDEX

Explanations

actions or commands given by an authority figure

oai_token-act-pair · gpt-3.5-turbo

instances of the word "ordered," particularly in contexts involving commands or directives

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

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

nown

-0.93

 sidx

-0.80

pmwiki

-0.77

uded

-0.76

facing

-0.75

rift

-0.74

common

-0.73

attery

-0.72

figure

-0.72

<?

-0.71

POSITIVE LOGITS

 evacuation

0.86

 demolition

0.85

 subordinates

0.83

 extermin

0.81

 destruction

0.80

 immediate

0.79

 extermination

0.77

 deletion

0.76

 removal

0.76

 halt

0.76

Activations Density 0.077%

actions or commands given by an authority figure

instances of the word "ordered," particularly in contexts involving commands or directives

No Comments

No Known Activations