INDEX

Explanations

the word "flash" at various activations

oai_token-act-pair · gpt-3.5-turbo

references to "flash."

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

employed

-0.79

lain

-0.76

thy

-0.74

nil

-0.74

mens

-0.73

avia

-0.71

 Guth

-0.71

 Kron

-0.68

 Cohn

-0.68

 Colo

-0.68

POSITIVE LOGITS

 flash

3.83

flash

2.87

Flash

2.45

 Flash

2.39

 flashes

2.28

 flashed

2.03

 flashing

1.94

 flashlight

1.55

 blink

1.46

 flashback

1.33

Activations Density 0.015%

the word "flash" at various activations

references to "flash."

No Comments

No Known Activations