INDEX

Explanations

instances of offense and apology in social interactions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 14-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L14R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.14.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

HORT

-0.15

 Municipal

-0.14

ë²Ķ

-0.14

loh

-0.14

ÐµÐ½Ð½Ñı

-0.14

chor

-0.14

annes

-0.13

ext

-0.13

_contrib

-0.13

DRV

-0.13

POSITIVE LOGITS

 offended

0.17

-sensitive

0.17

 sensitive

0.16

osaur

0.16

raya

0.15

ibal

0.15

 sensitivity

0.15

 sensit

0.15

 hurt

0.15

oux

0.15

Activations Density 0.247%

instances of offense and apology in social interactions

No Comments

No Known Activations