INDEX

Explanations

No Explanations Found

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

ocry

-0.68

 directories

-0.65

rency

-0.64

 rigging

-0.62

 disclaim

-0.60

 Aware

-0.60

Downloadha

-0.58

 Stamford

-0.57

 ellipt

-0.56

 Geoffrey

-0.56

POSITIVE LOGITS

MpServer

0.84

ãĥĥãĥī

0.83

trap

0.77

chini

0.76

tons

0.73

ãĥ³ãĤ¸

0.71

phia

0.71

0.69

¶ħ

0.67

IGHTS

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.