INDEX

Explanations

numbered sections or lists in a text

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

0:0.02

1:0.02

2:0.07

3:0.08

4:0.08

5:0.05

6:0.06

7:0.34

8:0.06

9:0.05

10:0.07

11:0.05

Negative Logits

 Cele

-1.57

 anus

-1.53

 Balt

-1.53

 refuge

-1.48

 cavity

-1.47

 haun

-1.47

 remnants

-1.45

 chew

-1.40

 saliva

-1.37

wre

-1.33

POSITIVE LOGITS

utics

1.64

Security

1.61

Understanding

1.60

engineering

1.59

QUIRE

1.58

custom

1.57

heit

1.51

hon

1.51

arning

1.50

 Understanding

1.49

Activations Density 0.001%

numbered sections or lists in a text

No Comments

No Known Activations