INDEX

Explanations

phrases indicating reasons or justifications for actions or beliefs

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

0:0.12

1:0.19

2:0.03

3:0.03

4:0.04

5:0.29

6:0.07

7:0.01

8:0.06

9:0.04

10:0.03

11:0.03

Negative Logits

UCT

-2.03

 captcha

-1.76

 bonded

-1.63

cel

-1.55

UC

-1.50

Document

-1.48

 condoms

-1.46

¶

-1.41

 parity

-1.40

uc

-1.40

POSITIVE LOGITS

nda

2.05

ses

2.00

dq

1.95

ongyang

1.90

 Yose

1.83

alion

1.82

tis

1.78

heny

1.76

sts

1.75

resa

1.74

Activations Density 0.005%

phrases indicating reasons or justifications for actions or beliefs

No Known Activations