INDEX

Explanations

comparisons or alternatives

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

0:0.10

1:0.03

2:0.17

3:0.09

4:0.09

5:0.07

6:0.04

7:0.02

8:0.09

9:0.14

10:0.07

11:0.03

Negative Logits

Congratulations

-1.36

armed

-1.18

Capt

-1.12

Step

-1.10

 Container

-1.07

 inspectors

-1.04

activated

-1.04

 Ranch

-1.00

 inexper

-1.00

liners

-0.99

POSITIVE LOGITS

ーク

1.18

imaru

1.16

cipl

1.12

ipolar

1.12

 reconcile

1.11

Lak

1.08

ographed

1.08

":"/

1.07

ONSORED

1.06

 mascul

1.04

Activations Density 0.005%

comparisons or alternatives

No Known Activations