INDEX

Explanations

punctuation and negative connotations related to spending and fiscal responsibility

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

0:0.01

1:0.02

2:0.04

3:0.04

4:0.04

5:0.03

6:0.47

7:0.09

8:0.03

9:0.05

10:0.07

11:0.06

Negative Logits

BB

-1.60

Cola

-1.52

Plot

-1.39

Opening

-1.38

Pool

-1.37

Match

-1.36

====

-1.32

Movie

-1.31

psc

-1.31

Asset

-1.28

POSITIVE LOGITS

rez

1.65

hest

1.50

schild

1.47

ctic

1.41

ileged

1.40

enged

1.29

undo

1.27

fetched

1.27

ritic

1.26

dden

1.24

Activations Density 0.013%

punctuation and negative connotations related to spending and fiscal responsibility

No Known Activations