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

 newcom

-0.73

 shelters

-0.68

 conflic

-0.67

 eleph

-0.63

azard

-0.63

 vain

-0.62

ALLY

-0.62

 corrid

-0.62

 Lauder

-0.61

FD

-0.60

POSITIVE LOGITS

requ

0.76

Reviewer

0.74

 Disciple

0.74

ruct

0.74

ratulations

0.73

lectic

0.70

pect

0.70

pection

0.68

 Seat

0.68

ptive

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.