INDEX

Explanations

topical reasons or justifications

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L2_Hcat_z_lr1.20e-03_l11.00e+00_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v4.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.2.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.11

1:0.03

2:0.08

3:0.21

4:0.04

5:0.08

6:0.04

7:0.03

8:0.10

9:0.11

10:0.07

11:0.03

Negative Logits

rocal

-1.28

orrow

-1.20

teness

-1.16

 Pattern

-1.13

kinson

-1.10

 omission

-1.08

rique

-1.06

tsy

-1.05

isoft

-1.05

avascript

-1.03

POSITIVE LOGITS

 achie

1.38

 Ever

1.31

 endorsements

1.16

Ever

1.15

CHA

1.14

 EVER

1.13

Favorite

1.13

 ranking

1.13

ever

1.11

 imaginable

1.11

Activations Density 0.102%

topical reasons or justifications

No Known Activations