INDEX

Explanations

terms related to hostile actions or entities

oai_token-act-pair · gpt-3.5-turbo

references to an "enemy."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

hots

-0.83

mberg

-0.80

auntlets

-0.79

razil

-0.78

-0.73

eret

-0.72

ucket

-0.71

oled

-0.71

akers

-0.70

regon

-0.70

POSITIVE LOGITS

 enemy

1.26

foe

1.14

 Enemy

1.13

 adversary

1.05

 combatants

0.97

enemy

0.95

 enemies

0.93

 adversaries

0.91

 opponent

0.88

 foes

0.87

Activations Density 0.011%

terms related to hostile actions or entities

references to an "enemy."

No Comments

No Known Activations