INDEX

Explanations

references to cover-ups regarding scandals or hidden information

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

2:0.05

3:0.06

4:0.16

5:0.03

6:0.03

7:0.34

8:0.05

9:0.03

10:0.08

11:0.07

Negative Logits

cho

-1.75

zai

-1.72

tera

-1.66

accompan

-1.65

comb

-1.62

quad

-1.49

KEY

-1.48

psc

-1.46

ombo

-1.46

Parameter

-1.44

POSITIVE LOGITS

 wrongdoing

2.27

 nonexistent

1.93

 illness

1.80

 innocence

1.74

 ignorance

1.66

 mistakes

1.65

 illnesses

1.65

 sickness

1.64

 Deaths

1.60

 doping

1.59

Activations Density 0.006%

references to cover-ups regarding scandals or hidden information

No Comments

No Known Activations