INDEX

Explanations

references to nuclear weapons and related policies

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

0:0.13

1:0.02

2:0.32

3:0.08

4:0.04

5:0.09

6:0.04

7:0.06

8:0.10

9:0.03

10:0.03

11:0.02

Negative Logits

 hust

-3.10

��

-2.94

 hospitality

-2.87

 Homeless

-2.78

 Huck

-2.73

 Yelp

-2.71

 Wage

-2.68

 homelessness

-2.65

 Volunte

-2.63

 Shoes

-2.63

POSITIVE LOGITS

nuclear

6.05

 nuclear

5.91

 Nuclear

5.85

 plutonium

5.54

uclear

5.28

 warheads

5.24

 atomic

4.82

 reactors

4.76

 uranium

4.67

iferation

4.58

Activations Density 0.338%

references to nuclear weapons and related policies

No Known Activations