INDEX

Explanations

No Explanations Found

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L9_Hcat_z_lr1.20e-03_l11.20e+00_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.9.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.08

1:0.08

2:0.08

3:0.08

4:0.07

5:0.09

6:0.07

7:0.09

8:0.09

9:0.07

10:0.07

11:0.08

Negative Logits

iewicz

-3.27

 Kenya

-3.03

 Croat

-3.03

Ser

-3.02

 Camer

-2.94

inelli

-2.94

 Interior

-2.93

 Latvia

-2.86

 Ethiopian

-2.85

 Issa

-2.83

POSITIVE LOGITS

weet

3.17

#$

2.79

yout

2.57

gh

2.50

 brainstorm

2.46

 unsub

2.45

 gigg

2.44

hig

2.40

ples

2.40

 catchy

2.39

Activations Density 0.000%

No Known Activations

This feature has no known activations.