INDEX

Explanations

instances of the word "look" and its variations used to direct attention

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

uner

-0.19

eka

-0.18

MENT

-0.17

eken

-0.17

ekt

-0.15

soever

-0.15

idor

-0.15

uctor

-0.14

/by

-0.14

ffen

-0.14

POSITIVE LOGITS

 closely

0.21

sharp

0.21

 familiar

0.21

ma

0.20

 Sharp

0.20

 sharp

0.19

outs

0.19

Fam

0.19

Ma

0.18

fant

0.18

Activations Density 0.017%

instances of the word "look" and its variations used to direct attention

No Comments

No Known Activations