INDEX

Explanations

phrases related to confidentiality and information control

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 16-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L16R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.16.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

 monot

-0.15

-0.14

 Hunger

-0.14

Wet

-0.14

-0.13

ëıħ

-0.13

arde

-0.13

pler

-0.13

.construct

-0.13

POSITIVE LOGITS

 sensitive

0.25

-sensitive

0.23

 sensitivity

0.19

Sensitive

0.19

 protection

0.19

ensitive

0.19

æķı

0.17

 protect

0.17

Rey

0.17

ä¿ĿæĬ¤

0.17

Activations Density 0.025%

phrases related to confidentiality and information control

No Comments

No Known Activations