INDEX

Explanations

the word "that."

oai_token-act-pair · gpt-4o-mini Triggered by @bot

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

Not in Any Lists

No Comments

Head Attr Weights

0:0.08

1:0.10

2:0.08

3:0.08

4:0.09

5:0.07

6:0.07

7:0.07

8:0.06

9:0.07

10:0.08

11:0.09

Negative Logits

 Downs

-2.00

 Bulgar

-1.79

 Guan

-1.74

 Tomb

-1.68

 Britann

-1.62

 monop

-1.59

lda

-1.58

vati

-1.57

ulia

-1.56

Lew

-1.53

POSITIVE LOGITS

 GOODMAN

1.94

oğ

1.75

soDeliveryDate

1.68

ividual

1.61

 certific

1.58

rentices

1.52

 adolesc

1.51

bringer

1.49

isode

1.47

 certify

1.46

Activations Density 0.000%

the word "that."

No Comments

No Known Activations