INDEX

Explanations

references to concepts of evil and immoral behavior

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 10-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L10R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.10.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

cerebras/SlimPajama-627B

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

posable

-0.18

-ing

-0.15

onic

-0.15

oon

-0.15

èį

-0.14

alaxy

-0.14

finity

-0.14

ils

-0.14

oons

-0.14

 controlId

-0.14

POSITIVE LOGITS

ous

0.18

ously

0.18

 deeds

0.17

uous

0.16

fully

0.16

lesi

0.16

 intent

0.15

TT

0.15

dest

0.15

-do

0.15

Activations Density 0.075%

references to concepts of evil and immoral behavior

No Comments

No Known Activations