INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 mole

-0.78

 Chaser

-0.73

 Petro

-0.70

Cot

-0.68

 Ares

-0.65

 shotguns

-0.64

Spy

-0.63

 Apollo

-0.62

Sul

-0.62

SOS

-0.62

POSITIVE LOGITS

anyahu

0.86

aligned

0.79

raltar

0.76

rients

0.74

eller

0.73

plementation

0.73

Ã±

0.69

Ç

0.68

grounds

0.67

Ã±a

0.66

Activations Density 0.000%

No Known Activations

This feature has no known activations.