INDEX

Explanations

references to the word "who" and its contextual variations related to identifying individuals

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

ted

-0.17

bian

-0.17

mente

-0.17

scan

-0.16

ting

-0.16

pure

-0.15

orial

-0.15

cline

-0.15

ned

-0.15

robat

-0.15

POSITIVE LOGITS

 else

0.42

 exactly

0.33

/how

0.30

 ELSE

0.27

 Else

0.25

else

0.24

 amongst

0.24

ever

0.24

ops

0.23

 among

0.23

Activations Density 0.038%

references to the word "who" and its contextual variations related to identifying individuals

No Comments

No Known Activations