INDEX

Explanations

specific statements and quotations related to investigative processes or community responses

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

0:0.06

1:0.40

2:0.05

3:0.04

4:0.07

5:0.02

6:0.06

7:0.08

8:0.03

9:0.04

10:0.04

11:0.07

Negative Logits

orno

-2.74

 Thumbnails

-2.69

-2.55

minist

-2.42

sonian

-2.39

 cropped

-2.37

Hol

-2.32

borough

-2.29

anny

-2.27

 hipp

-2.27

POSITIVE LOGITS

Strong

8.32

 Strong

7.98

 weak

7.42

 strong

7.31

 Weak

7.13

strong

7.04

Weak

6.86

 weaker

6.66

weak

6.54

 stronger

6.51

Activations Density 0.093%

specific statements and quotations related to investigative processes or community responses

No Known Activations