INDEX

Explanations

instances of the number "one" being emphasized or contrasted with something else

oai_token-act-pair · gpt-3.5-turbo

phrases indicating the concept of pairs or groups of two

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

sed

-0.66

 Shutterstock

-0.63

uel

-0.62

odor

-0.62

proc

-0.60

å½

-0.59

æ³

-0.59

services

-0.57

ury

-0.57

phies

-0.57

POSITIVE LOGITS

 halves

0.94

ngth

0.71

auga

0.70

 accuser

0.68

 iteration

0.66

 side

0.64

 apiece

0.60

 accus

0.59

here

0.59

yden

0.59

Activations Density 0.086%

instances of the number "one" being emphasized or contrasted with something else

phrases indicating the concept of pairs or groups of two

No Comments

No Known Activations