INDEX

Explanations

references to evil and malice

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 6-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L6R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.6.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

-0.17

etch

-0.16

ekk

-0.15

-issue

-0.15

à¤²à¤¬

-0.14

åĭĻ

-0.14

nett

-0.14

luvÃŃ

-0.14

dlg

-0.14

_phrase

-0.13

POSITIVE LOGITS

 deeds

0.16

ÛĮÙĩ

0.15

ously

0.15

fully

0.15

ous

0.15

rea

0.15

lesi

0.14

á»ģ

0.14

rd

0.14

 Carp

0.14

Activations Density 0.022%

references to evil and malice

No Comments

No Known Activations