INDEX

Explanations

statements regarding legal proceedings and law enforcement actions

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

0:0.06

1:0.02

2:0.07

3:0.03

4:0.06

5:0.04

6:0.23

7:0.05

8:0.06

9:0.24

10:0.03

11:0.06

Negative Logits

 Ashes

-3.78

ICE

-3.71

ulhu

-3.64

 sulph

-3.50

ails

-3.38

die

-3.38

 Blizz

-3.38

 aval

-3.36

 Dota

-3.35

steam

-3.32

POSITIVE LOGITS

 Patterson

10.54

terson

6.06

Pat

5.82

Pat

5.20

pat

4.82

 Patt

4.79

 Pearson

4.73

patrick

4.71

 Perkins

4.69

 Pett

4.58

Activations Density 0.002%

statements regarding legal proceedings and law enforcement actions

No Known Activations