INDEX

Explanations

references to actions or events happening behind closed doors

oai_token-act-pair · gpt-3.5-turbo

references to doors, especially in contexts of closure, opening, and metaphorical significance

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

TY

-0.80

TING

-0.76

amera

-0.70

ting

-0.69

ency

-0.68

TX

-0.68

CTV

-0.68

allah

-0.66

 Astro

-0.63

OTAL

-0.63

POSITIVE LOGITS

doors

1.08

 doors

1.06

holes

1.01

mith

0.97

door

0.94

pring

0.92

hips

0.86

 Doors

0.85

hole

0.85

bell

0.81

Activations Density 0.010%

references to actions or events happening behind closed doors

references to doors, especially in contexts of closure, opening, and metaphorical significance

No Comments

No Known Activations