INDEX

Explanations

references to specific individuals, such as names or identifiers

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

0:0.04

1:0.03

2:0.03

3:0.07

4:0.02

5:0.03

6:0.17

7:0.03

8:0.03

9:0.45

10:0.02

11:0.02

Negative Logits

Roberts

-3.81

Iv

-3.71

Jackson

-3.59

Rob

-3.44

rob

-3.43

Jordan

-3.43

 Carlson

-3.37

 Bree

-3.35

Carl

-3.34

Raf

-3.32

POSITIVE LOGITS

um

6.96

UM

6.67

Gum

6.15

Dum

6.08

ums

5.70

Lum

5.47

umm

5.21

Kum

5.07

Mum

4.90

umo

4.89

Activations Density 0.007%

references to specific individuals, such as names or identifiers

No Known Activations