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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

terday

-0.79

 mileage

-0.66

 Lavrov

-0.64

 Yanuk

-0.64

usk

-0.64

 annexation

-0.64

 bloodshed

-0.62

ICLE

-0.61

 setbacks

-0.61

OTO

-0.59

POSITIVE LOGITS

hid

0.84

odied

0.78

 actionGroup

0.73

ajo

0.73

rift

0.72

oop

0.66

habi

0.64

 oppos

0.63

Charg

0.63

flush

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.