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

enz

-0.86

VICE

-0.79

ERROR

-0.77

Deal

-0.73

TY

-0.73

HI

-0.72

ECH

-0.71

help

-0.71

ADS

-0.71

ITS

-0.69

POSITIVE LOGITS

 dracon

0.70

htt

0.67

Goo

0.67

 mete

0.66

 manif

0.65

tho

0.64

 confir

0.64

 Liberties

0.63

 veter

0.63

 streng

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.