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

izen

-0.86

Fake

-0.73

Bey

-0.68

 Migration

-0.68

izens

-0.66

Discover

-0.66

ãĥ¼ãĥĨãĤ£

-0.66

Rules

-0.65

reme

-0.64

ãĥ¼ãĥĨ

-0.64

POSITIVE LOGITS

dq

0.91

 corps

0.73

 profession

0.65

 crest

0.62

cav

0.62

 mortg

0.61

hood

0.59

 backbone

0.59

 plunder

0.59

iggins

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.