INDEX

Explanations

instances of the word "off."

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L10_Hcat_z_lr1.20e-03_l11.30e+00_ds24576_bs4096_dc1.00e-05_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.10.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.08

1:0.07

2:0.08

3:0.08

4:0.09

5:0.07

6:0.07

7:0.08

8:0.09

9:0.08

10:0.08

11:0.08

Negative Logits

ilater

-2.70

 Afric

-2.69

Sham

-2.67

 initi

-2.62

 proofs

-2.56

 seals

-2.46

光

-2.46

 Nicaragua

-2.37

 Atlantis

-2.31

ryu

-2.29

POSITIVE LOGITS

VB

3.86

GHz

2.57

CHA

2.56

apper

2.47

iT

2.41

patch

2.39

 Reincarnated

2.35

2.34

/+

2.32

 Clojure

2.31

Activations Density 0.000%

instances of the word "off."

No Known Activations