INDEX

Explanations

references to actions or events involving expectations, outcomes, or evaluations

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L3_Hcat_z_lr1.20e-03_l19.00e-01_ds24576_bs4096_dc1.00e-06_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.3.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.04

1:0.08

2:0.10

3:0.02

4:0.04

5:0.07

6:0.22

7:0.13

8:0.08

9:0.07

10:0.06

11:0.05

Negative Logits

cation

-1.10

gery

-1.07

ilings

-1.06

iuses

-1.05

cerning

-1.01

cus

-1.00

gements

-0.98

itus

-0.97

isphere

-0.97

../

-0.97

POSITIVE LOGITS

 Cosponsors

1.32

フォ

1.05

NK

1.03

 however

1.03

BIL

1.02

 therefore

1.01

atche

0.96

Yad

0.95

 acronym

0.93

NP

0.92

Activations Density 0.180%

references to actions or events involving expectations, outcomes, or evaluations

No Known Activations