INDEX

Explanations

phrases related to recording or capturing visual imagery such as "filmed" and "photographed"

oai_token-act-pair · gpt-3.5-turbo

references to filming and video production

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

Weak

-0.70

 utility

-0.66

ãĤ§

-0.65

 Advantage

-0.63

 stability

-0.63

ulner

-0.62

 default

-0.61

 disposition

-0.59

 inheritance

-0.59

Atk

-0.59

POSITIVE LOGITS

 filmed

3.54

 photographed

2.46

 recorded

1.80

 filming

1.79

 taped

1.74

 videot

1.72

recorded

1.45

 aired

1.41

 televised

1.40

 streamed

1.38

Activations Density 0.016%

phrases related to recording or capturing visual imagery such as "filmed" and "photographed"

references to filming and video production

No Comments

No Known Activations