INDEX

Explanations

filler words or interjections (such as "um" or "uh")

oai_token-act-pair · gpt-3.5-turbo

filler words or hesitation markers in speech

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

hips

-0.85

Constructed

-0.72

BOOK

-0.70

ingham

-0.66

Kings

-0.66

 rendered

-0.63

icipated

-0.63

ebook

-0.63

Adds

-0.62

hire

-0.62

POSITIVE LOGITS

pty

0.92

ugh

0.88

br

0.83

ming

0.80

rah

0.80

uge

0.76

bers

0.76

amoto

0.75

othe

0.75

yssey

0.75

Activations Density 0.014%

filler words or interjections (such as "um" or "uh")

filler words or hesitation markers in speech

No Comments

No Known Activations