INDEX

Explanations

references to media reporting and public relations activity

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

1:0.03

2:0.04

3:0.04

4:0.36

5:0.05

6:0.07

7:0.07

8:0.01

9:0.15

10:0.02

11:0.07

Negative Logits

�

-3.05

requires

-2.91

amac

-2.75

ntil

-2.69

orest

-2.67

upper

-2.63

ajor

-2.59

ht

-2.57

��

-2.55

father

-2.55

POSITIVE LOGITS

 publicity

5.40

 sensational

4.68

 portray

4.67

 disinformation

4.40

 tabloid

4.37

 media

4.18

 headlines

4.17

 hype

4.16

 propaganda

4.10

 journalists

4.09

Activations Density 0.572%

references to media reporting and public relations activity

No Known Activations