INDEX

Explanations

phrases related to ethical or legal issues involving misconduct

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 21-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L21R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.21.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

cerebras/SlimPajama-627B

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 upstairs

-0.20

ä¸Ĭ

-0.20

ä¸ĬçļĦ

-0.19

north

-0.18

åĮĹ

-0.18

 north

-0.18

 ä¸Ĭ

-0.18

é«ĺ

-0.17

æľĢé«ĺ

-0.17

ä¸ĬäºĨ

-0.17

POSITIVE LOGITS

 lower

0.65

 bottom

0.65

 below

0.65

ä¸ĭ

0.60

below

0.57

bottom

0.56

 Below

0.56

 ä¸ĭ

0.55

 Bottom

0.55

 lowest

0.55

Activations Density 0.361%

phrases related to ethical or legal issues involving misconduct

No Comments

No Known Activations