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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

velt

-0.94

 Regions

-0.83

 Rooms

-0.81

 Roads

-0.75

itiveness

-0.71

 Society

-0.71

ousy

-0.70

iquette

-0.69

 leaflets

-0.69

 Streets

-0.68

POSITIVE LOGITS

 vested

0.68

 streng

0.62

 deaf

0.62

aged

0.62

 hiber

0.62

 Michaels

0.62

 sled

0.60

 swallowed

0.60

pron

0.60

 knit

0.58

Activations Density 0.000%

No Known Activations

This feature has no known activations.