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

edes

-0.76

 retard

-0.72

eree

-0.65

isers

-0.62

 Spock

-0.60

antics

-0.60

 blinded

-0.59

actor

-0.58

 Canaver

-0.57

izers

-0.57

POSITIVE LOGITS

 conflic

0.83

ICAN

0.70

ternity

0.70

olit

0.70

cil

0.69

 proport

0.69

Reloaded

0.67

repair

0.66

 occas

0.65

prus

0.65

Activations Density 0.000%

No Known Activations

This feature has no known activations.