INDEX

Explanations

references to historical positions and their holders

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 26-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L26R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.26.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

ahat

-0.16

ollen

-0.14

ornings

-0.14

noh

-0.14

ierten

-0.13

inski

-0.13

Ð»Ñİ

-0.13

å¢

-0.13

 themselves

-0.13

endor

-0.13

POSITIVE LOGITS

 appointment

0.28

 succeed

0.25

 succeeds

0.23

åħ¼

0.23

 resign

0.23

 appointments

0.23

 succeeding

0.22

 succeeded

0.21

appointment

0.21

 vacancy

0.21

Activations Density 0.110%

references to historical positions and their holders

No Comments

No Known Activations