INDEX

Explanations

references to specific places or cities

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.10

1:0.32

2:0.05

3:0.05

4:0.04

5:0.07

6:0.04

7:0.02

8:0.09

9:0.07

10:0.05

11:0.04

Negative Logits

owell

-1.91

button

-1.81

ABLE

-1.79

LOG

-1.70

adian

-1.67

 crow

-1.61

udder

-1.58

addle

-1.57

tackle

-1.56

ebook

-1.55

POSITIVE LOGITS

San

2.56

San

2.44

 Sanford

2.18

 Sant

2.03

 Gren

2.02

san

1.98

 Santiago

1.98

angan

1.95

 Franch

1.95

 Marina

1.94

Activations Density 0.001%

references to specific places or cities

No Known Activations