INDEX

Explanations

phrases indicating a strong warning or caution

oai_token-act-pair · gpt-3.5-turbo

instances of the word "warn" or its variations

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

ota

-0.68

 Esports

-0.65

 transc

-0.60

 participation

-0.60

taking

-0.59

Bou

-0.57

 embodied

-0.57

 acquired

-0.57

RG

-0.56

 Participant

-0.56

POSITIVE LOGITS

 warn

3.76

 warns

2.19

 warning

1.99

warn

1.99

 warnings

1.91

 warned

1.88

 Warn

1.85

warning

1.64

 caution

1.60

 WARN

1.55

Activations Density 0.014%

phrases indicating a strong warning or caution

instances of the word "warn" or its variations

No Comments

No Known Activations