INDEX

Explanations

actions related to arranging or organizing events or objects

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L2_Hcat_z_lr1.20e-03_l11.00e+00_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v4.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.2.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

1:0.02

2:0.11

3:0.06

4:0.32

5:0.05

6:0.02

7:0.02

8:0.13

9:0.11

10:0.04

11:0.02

Negative Logits

 whiff

-1.40

famous

-1.40

 Rapids

-1.35

 Norse

-1.34

 Heisman

-1.31

brew

-1.29

baugh

-1.29

 Frazier

-1.26

 Lerner

-1.23

young

-1.23

POSITIVE LOGITS

 arrang

1.80

 ACTIONS

1.70

IFE

1.65

wed

1.51

adia

1.47

女

1.46

APE

1.44

arenthood

1.44

Gameplay

1.43

OLOG

1.43

Activations Density 0.004%

actions related to arranging or organizing events or objects

No Known Activations