INDEX

Explanations

legal terminology and references to sentencing, charges, and offenses

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

0:0.06

1:0.09

2:0.06

3:0.09

4:0.10

5:0.11

6:0.06

7:0.09

8:0.09

9:0.06

10:0.06

11:0.07

Negative Logits

eport

-2.59

 memos

-2.59

Indiana

-2.31

Washington

-2.21

 lately

-2.20

CLS

-2.17

 casinos

-2.14

conom

-2.12

 logs

-2.10

 abuses

-2.09

POSITIVE LOGITS

£

3.32

 Firstly

3.00

 Whilst

2.96

 Tube

2.82

 Flavoring

2.81

 whilst

2.80

£

2.80

 UKIP

2.68

NZ

2.65

 Midlands

2.64

Activations Density 0.007%

legal terminology and references to sentencing, charges, and offenses

No Known Activations