INDEX

Explanations

the word "so" in various contexts, indicating a focus on expressions of emphasis or transition

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

so

-0.21

work

-0.19

so

-0.19

ting

-0.18

gram

-0.18

st

-0.17

tr

-0.17

stand

-0.17

kill

-0.16

POSITIVE LOGITS

-called

0.43

ooo

0.29

oooo

0.28

apy

0.26

ething

0.25

iled

0.25

 forth

0.24

oooooooo

0.24

oth

0.24

oner

0.24

Activations Density 0.042%

the word "so" in various contexts, indicating a focus on expressions of emphasis or transition

No Comments

No Known Activations