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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

itte

-0.74

-0.72

 infiltration

-0.66

monary

-0.64

PLA

-0.63

 trends

-0.63

itual

-0.61

igraph

-0.61

 infiltrate

-0.60

 withd

-0.60

POSITIVE LOGITS

 Drag

0.73

matter

0.66

 Jagu

0.66

 Ulster

0.65

urance

0.63

aus

0.63

 Zimmer

0.63

 Indianapolis

0.63

 Cele

0.62

 Baltimore

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.