INDEX

Explanations

numeric values representing a maximum limit or capacity

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate limitations or thresholds

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

bane

-0.88

FORE

-0.83

perse

-0.81

guy

-0.77

 Guys

-0.76

hell

-0.76

ãĥīãĥ©

-0.73

velt

-0.72

keeper

-0.68

vor

-0.68

POSITIVE LOGITS

 allowable

1.08

 amount

1.04

 extent

0.97

 likelihood

0.87

 punishment

0.81

imize

0.81

egal

0.80

nesota

0.80

 maximum

0.79

 amounts

0.79

Activations Density 0.013%

numeric values representing a maximum limit or capacity

phrases that indicate limitations or thresholds

No Comments

No Known Activations