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

ngth

-0.96

bris

-0.84

 toget

-0.82

 tremend

-0.82

ouble

-0.81

sonian

-0.79

ogun

-0.74

elight

-0.74

hetti

-0.73

sung

-0.73

POSITIVE LOGITS

 alcoholic

0.76

 Ares

0.71

 Authorization

0.70

 undead

0.68

 Palin

0.66

 MDMA

0.66

 terms

0.65

 undocumented

0.65

FY

0.64

 illegal

0.63

Activations Density 0.000%

No Known Activations

This feature has no known activations.