INDEX

Explanations

phrases indicating instructions or commands

oai_token-act-pair · gpt-3.5-turbo

instructions or recommendations to utilize specific tools or methods

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

 undertaking

-0.71

SPONSORED

-0.66

 confer

-0.65

 Qiao

-0.64

 indebted

-0.63

 unfit

-0.63

 behold

-0.62

Ń·

-0.60

 undergoing

-0.59

Î½

-0.59

POSITIVE LOGITS

FUL

1.13

fully

1.08

full

1.06

fulness

0.97

ful

0.92

 Cases

0.78

 shortcuts

0.75

 condoms

0.74

 whichever

0.73

 sparing

0.71

Activations Density 0.092%

phrases indicating instructions or commands

instructions or recommendations to utilize specific tools or methods

No Comments

No Known Activations