INDEX

Explanations

phrases related to military and defense

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

colored

-0.83

atl

-0.76

 localized

-0.75

NetMessage

-0.73

 taco

-0.71

 modeled

-0.71

 colored

-0.68

 wired

-0.66

 loop

-0.66

 color

-0.65

POSITIVE LOGITS

 Defence

4.11

 defence

2.67

 Defense

2.28

 defences

1.79

Defense

1.70

 Nato

1.44

 armoured

1.38

 Honour

1.36

 DEFENSE

1.35

 defense

1.30

Activations Density 0.008%

phrases related to military and defense

No Comments

No Known Activations