INDEX

Explanations

phrases requesting actions or conveying politeness

oai_token-act-pair · gpt-3.5-turbo

requests for action or feedback

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

arc

-0.77

ARC

-0.76

advertisement

-0.73

ription

-0.69

senal

-0.68

existence

-0.67

ula

-0.64

phrine

-0.63

mund

-0.61

bang

-0.60

POSITIVE LOGITS

sir

0.92

 Ignore

0.80

 fill

0.78

 pardon

0.77

 ignore

0.76

 inquire

0.74

 excuse

0.74

 forgive

0.73

 beware

0.72

ignore

0.71

Activations Density 0.013%

phrases requesting actions or conveying politeness

requests for action or feedback

No Comments

No Known Activations