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

 behav

-0.71

 Dynamics

-0.71

abouts

-0.70

kef

-0.68

 Decay

-0.66

 Palestin

-0.65

 profession

-0.64

 describ

-0.60

 contrace

-0.60

dp

-0.60

POSITIVE LOGITS

è»

0.69

assembly

0.67

RAFT

0.65

isance

0.65

oteric

0.63

AMY

0.63

cyclop

0.62

catentry

0.62

 Hots

0.62

ersive

0.62

Activations Density 0.000%

No Known Activations

This feature has no known activations.