INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 challeng

-0.76

erous

-0.75

 prosec

-0.68

 advoc

-0.67

 welf

-0.67

conn

-0.66

 subcontract

-0.66

 detrim

-0.66

vex

-0.66

 prosecute

-0.65

POSITIVE LOGITS

encer

0.75

Prince

0.64

 Hein

0.62

 SOLD

0.62

interstitial

0.61

 presence

0.60

Sho

0.60

 thick

0.60

usky

0.59

Hik

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.