INDEX

Explanations

conjunctions used to connect clauses or phrases

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

1:0.09

2:0.08

3:0.08

4:0.09

5:0.08

6:0.08

7:0.08

8:0.08

9:0.06

10:0.06

11:0.08

Negative Logits

ffen

-2.77

 Mariners

-2.73

astery

-2.63

 Castro

-2.54

 embassy

-2.52

 Rubio

-2.49

xon

-2.46

jam

-2.40

ham

-2.40

apple

-2.38

POSITIVE LOGITS

 Consent

2.81

ENE

2.79

robe

2.77

Form

2.71

uyomi

2.51

 subscript

2.50

 Forms

2.49

 Solitaire

2.45

 catalyst

2.33

 unknow

2.30

Activations Density 0.000%

conjunctions used to connect clauses or phrases

No Known Activations