INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L3_Hcat_z_lr1.20e-03_l19.00e-01_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.3.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.09

1:0.08

2:0.07

3:0.05

4:0.10

5:0.07

6:0.09

7:0.10

8:0.08

9:0.07

10:0.07

11:0.09

Negative Logits

 contrace

-1.71

 bench

-1.42

 mastered

-1.38

 pregn

-1.35

cov

-1.32

 clipboard

-1.32

piv

-1.30

 crappy

-1.29

 curs

-1.29

 scrim

-1.28

POSITIVE LOGITS

runners

1.42

Anth

1.41

Bir

1.34

NET

1.34

hua

1.30

 Esports

1.29

Ca

1.28

Redditor

1.26

Hor

1.22

 Boxing

1.22

Activations Density 0.000%

No Known Activations

This feature has no known activations.