INDEX

Explanations

expressions of confusion or lack of understanding

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.02

1:0.01

2:0.09

3:0.09

4:0.19

5:0.03

6:0.08

7:0.27

8:0.03

9:0.04

10:0.05

11:0.06

Negative Logits

oided

-1.86

 rall

-1.74

uilding

-1.51

sembly

-1.51

aughtered

-1.42

Fall

-1.42

ework

-1.40

 cleanup

-1.39

eday

-1.38

eworks

-1.37

POSITIVE LOGITS

 anymore

2.06

 slightest

1.51

 Argument

1.43

 nowadays

1.41

 jokes

1.38

 mention

1.36

 firsthand

1.33

nor

1.30

 disappro

1.30

utters

1.30

Activations Density 0.009%

expressions of confusion or lack of understanding

No Known Activations