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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ailability

-0.84

mir

-0.80

interstitial

-0.79

hur

-0.65

ometown

-0.62

Correction

-0.61

agara

-0.61

agher

-0.60

ingham

-0.59

romeda

-0.58

POSITIVE LOGITS

Zone

0.72

 Lobby

0.71

Soc

0.62

Boss

0.61

Commerce

0.61

 lobb

0.60

 zones

0.59

Bloom

0.59

Handle

0.59

 evenings

0.59

Activations Density 0.000%

No Known Activations

This feature has no known activations.