INDEX

Explanations

references to the internet or online content

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L10_Hcat_z_lr1.20e-03_l11.30e+00_ds24576_bs4096_dc1.00e-05_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.10.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

2:0.07

3:0.07

4:0.07

5:0.09

6:0.08

7:0.08

8:0.08

9:0.07

10:0.09

11:0.07

Negative Logits

emic

-2.73

 Underworld

-2.63

adium

-2.33

icidal

-2.32

 Sanctuary

-2.32

Assembly

-2.30

ucc

-2.25

 Shutdown

-2.22

 Wicked

-2.21

vine

-2.21

POSITIVE LOGITS

izabeth

2.70

Iv

2.67

 Pesh

2.57

 Euros

2.53

 UNESCO

2.53

 Ottoman

2.46

Io

2.44

 Picard

2.43

 Estonia

2.43

 Mons

2.40

Activations Density 0.000%

references to the internet or online content

No Known Activations