INDEX

Explanations

references to significant sporting achievements and accolades

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

0:0.16

1:0.02

2:0.10

3:0.06

4:0.03

5:0.04

6:0.03

7:0.04

8:0.39

9:0.02

10:0.03

11:0.01

Negative Logits

aneers

-3.97

 Skyrim

-3.73

 Bills

-3.45

 Homs

-3.44

YPG

-3.38

Syrian

-3.37

ersive

-3.36

 Leia

-3.27

olkien

-3.23

yrim

-3.21

POSITIVE LOGITS

 Tennis

6.15

 tennis

5.85

ATP

5.62

Nad

4.56

bledon

4.54

Wim

4.51

 Slam

4.21

 disqual

4.02

 tournaments

3.96

 Sere

3.91

Activations Density 0.053%

references to significant sporting achievements and accolades

No Known Activations