INDEX

Explanations

instances of the word "think" and its variations used to prompt reflection or consideration

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 6-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L6R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.6.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

ãĤ¸ãĤ¢

-0.16

ious

-0.16

leton

-0.14

vido

-0.14

ldr

-0.14

hle

-0.14

ÏĨÏħ

-0.13

iano

-0.13

Ð°ÑĢÑĩ

-0.13

Ø±ÙĪÙħ

-0.13

POSITIVE LOGITS

tors

0.16

ock

0.15

æĭ¥

0.15

amina

0.15

erve

0.14

orthand

0.14

ascar

0.14

LOAT

0.14

lÃ¼

0.13

ollo

0.13

Activations Density 0.024%

instances of the word "think" and its variations used to prompt reflection or consideration

No Comments

No Known Activations