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

udos

-0.77

atar

-0.73

iman

-0.72

anos

-0.71

achel

-0.69

Reviewed

-0.68

 Patton

-0.68

AMA

-0.67

hardt

-0.63

ANC

-0.63

POSITIVE LOGITS

 ACTIONS

0.65

 reactive

0.64

 entropy

0.64

 actionGroup

0.63

paio

0.61

 clot

0.60

 membr

0.60

 tremend

0.59

 lear

0.59

 manif

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.