INDEX

Explanations

references to editorial submissions or contact information for publications

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

0:0.03

1:0.02

2:0.02

3:0.05

4:0.10

5:0.25

6:0.03

7:0.04

8:0.14

9:0.22

10:0.02

11:0.02

Negative Logits

tho

-2.98

theless

-2.93

 whilst

-2.67

 realise

-2.53

emale

-2.48

)?

-2.41

;;;;;;;;;;;;

-2.40

 alot

-2.40

etheless

-2.32

terday

-2.29

POSITIVE LOGITS

cling

2.32

angering

2.21

rup

2.19

 empowering

2.15

 issuing

2.07

aga

2.04

2.02

 Berm

2.02

overe

2.01

 offending

1.99

Activations Density 0.002%

references to editorial submissions or contact information for publications

No Known Activations