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

bol

-0.79

 Masquerade

-0.72

 Sacrifice

-0.68

 Slate

-0.65

 Bravo

-0.65

mination

-0.64

 Bryce

-0.64

 slate

-0.62

 Americas

-0.62

 Confederacy

-0.59

POSITIVE LOGITS

igham

1.01

artney

0.90

rehend

0.83

uden

0.83

urn

0.82

izzard

0.81

liam

0.81

eals

0.80

akura

0.77

 sidx

0.77

Activations Density 0.000%

No Known Activations

This feature has no known activations.