INDEX

Explanations

No Explanations Found

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

IFrame

Link

Negative Logits

MP

-0.81

abad

-0.73

 Stones

-0.71

acies

-0.70

SpaceEngineers

-0.69

rail

-0.66

 Course

-0.64

ulty

-0.64

airs

-0.64

oric

-0.63

POSITIVE LOGITS

 masturb

0.71

inki

0.67

Subscribe

0.67

ipolar

0.66

det

0.65

lia

0.64

pid

0.62

 orgasm

0.61

 welf

0.61

___

0.61

Activations Density 0.000%

No Known Activations

This feature has no known activations.