INDEX

Explanations

repeated instances of the word "the" and similar articles or pronouns

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

0:0.04

1:0.02

2:0.11

3:0.07

4:0.15

5:0.04

6:0.08

7:0.14

8:0.04

9:0.05

10:0.06

11:0.14

Negative Logits

 precaution

-1.48

 objection

-1.47

 suspicions

-1.47

 retali

-1.45

empt

-1.41

 alarms

-1.41

 outright

-1.39

 fals

-1.38

 reins

-1.38

 flagged

-1.38

POSITIVE LOGITS

 handsome

1.54

 spiritually

1.50

 Serving

1.50

ardless

1.49

 greatness

1.48

 beautiful

1.45

thood

1.40

olars

1.39

 rich

1.39

装

1.38

Activations Density 0.000%

repeated instances of the word "the" and similar articles or pronouns

No Known Activations