INDEX

Explanations

the inclusion of the phrase "how to" indicating instructional content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 2-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L2R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.2.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

uem

-0.15

alsa

-0.15

lys

-0.14

 Initiative

-0.14

quis

-0.13

otive

-0.13

 Fern

-0.13

ìķħ

-0.13

.am

-0.13

ffm

-0.13

POSITIVE LOGITS

cone

0.17

arton

0.14

ìĦ

0.14

 Sherman

0.14

rone

0.14

obus

0.14

 NDEBUG

0.13

Ð°Ð·

0.13

 stagn

0.13

_GUID

0.13

Activations Density 0.021%

the inclusion of the phrase "how to" indicating instructional content

No Comments

No Known Activations