INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

illery

-0.67

Klu

-0.66

 wrists

-0.65

ye

-0.65

iolet

-0.63

access

-0.62

 wrist

-0.62

omore

-0.62

 Nobel

-0.61

 Hort

-0.58

POSITIVE LOGITS

abouts

0.78

 corrid

0.76

ctuary

0.72

vironment

0.72

ãĤ¦ãĤ¹

0.70

Dynamic

0.67

VERTISEMENT

0.67

Tracker

0.65

 persuaded

0.64

rompt

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.