INDEX

Explanations

phrases emphasizing the word "most" and quantifying a majority

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 29-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L29R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.29.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

agli

-0.15

AffineTransform

-0.15

combe

-0.14

åĲĦç§į

-0.14

ê·¹

-0.14

 Various

-0.14

inue

-0.14

ãĢģãĢģ

-0.14

inki

-0.13

POSITIVE LOGITS

 portion

0.31

/all

0.26

 part

0.25

 parte

0.25

portion

0.23

 portions

0.23

 chunk

0.22

 share

0.19

 ÑĩÐ°ÑģÑĤÑĮ

0.18

 parts

0.18

Activations Density 0.056%

phrases emphasizing the word "most" and quantifying a majority

No Comments

No Known Activations