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

ubi

-0.75

odcast

-0.73

ument

-0.73

ided

-0.71

ound

-0.66

 Platform

-0.62

Ĥİ

-0.62

Machine

-0.61

 Towards

-0.61

idable

-0.61

POSITIVE LOGITS

 glim

0.88

ilyn

0.81

senal

0.80

 philos

0.73

 citiz

0.70

Cly

0.70

 Rhodes

0.70

edIn

0.69

 undermin

0.68

 swaps

0.67

Activations Density 0.000%

No Known Activations

This feature has no known activations.