INDEX

Explanations

the word "which" and its variations, indicating a focus on choices or inquiries

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 26-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L26R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.26.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

sson

-0.16

loff

-0.15

ajo

-0.14

orig

-0.14

 outset

-0.14

 behalf

-0.13

til

-0.13

empo

-0.13

orial

-0.13

POSITIVE LOGITS

 ones

0.31

soever

0.29

-ever

0.28

 direction

0.26

ones

0.24

of

0.24

/how

0.23

 Ð¸Ð¼ÐµÐ½Ð½Ð¾

0.23

 Ones

0.20

 Wich

0.20

Activations Density 0.037%

the word "which" and its variations, indicating a focus on choices or inquiries

No Comments

No Known Activations