INDEX

Explanations

words related to hostility or conflict, particularly with an intensity implying adversaries or opponents

oai_token-act-pair · gpt-3.5-turbo

references to adversarial groups or individuals

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Ħ¢

-0.82

printed

-0.77

electric

-0.74

oled

-0.69

oard

-0.69

oret

-0.67

auntlets

-0.65

 Cout

-0.65

 Jackets

-0.64

eret

-0.63

POSITIVE LOGITS

foe

1.14

 Enemy

1.03

 enemy

1.03

 adversaries

1.00

 enemies

0.97

hip

0.90

 foes

0.90

 undermin

0.88

 Enemies

0.88

 adversary

0.88

Activations Density 0.019%

words related to hostility or conflict, particularly with an intensity implying adversaries or opponents

references to adversarial groups or individuals

No Comments

No Known Activations