INDEX

Explanations

strong affirmations or denials in discourse

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

0:0.03

1:0.03

2:0.09

3:0.06

4:0.01

5:0.04

6:0.15

7:0.07

8:0.04

9:0.04

10:0.08

11:0.30

Negative Logits

 Odyssey

-1.12

 nearest

-1.10

 nearby

-1.09

 Crusade

-1.05

pron

-1.02

lator

-1.00

 monthly

-0.99

 roughly

-0.99

 Paste

-0.98

 Summit

-0.97

POSITIVE LOGITS

orsche

1.28

omical

1.19

enez

1.16

rists

1.15

istas

1.09

 calib

1.08

WHERE

1.06

osite

1.06

ه

1.05

oso

1.05

Activations Density 0.025%

strong affirmations or denials in discourse

No Known Activations