INDEX

Explanations

phrases related to targeted actions or attacks

oai_token-act-pair · gpt-3.5-turbo

references to targeting or targeting strategies in a context involving military or cyber operations

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

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 trem

-0.67

 vintage

-0.66

 slide

-0.66

emb

-0.66

 bubble

-0.66

 Auditor

-0.65

Day

-0.64

 chapter

-0.64

rot

-0.62

 plunge

-0.62

POSITIVE LOGITS

target

2.62

second

2.04

points

1.63

match

1.41

ranked

1.40

default

1.33

seconds

1.31

cur

1.25

local

1.23

hitting

1.21

Activations Density 0.039%

phrases related to targeted actions or attacks

references to targeting or targeting strategies in a context involving military or cyber operations

No Comments

No Known Activations