INDEX

Explanations

sections that feature ranked or categorized lists, particularly in a top ten or top list format

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 3-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L3R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.3.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

agher

-0.16

agar

-0.16

arrants

-0.15

ment

-0.15

 Fork

-0.14

bane

-0.14

umph

-0.14

Sad

-0.14

 Grammar

-0.13

POSITIVE LOGITS

gil

0.19

 Ø±Ø³Ùħ

0.15

erten

0.15

 Ð¿ÐµÑĢÐµÐ±

0.15

 Unidos

0.14

kur

0.14

oplay

0.14

asmus

0.14

<Guid

0.14

yÃ¼k

0.14

Activations Density 0.012%

sections that feature ranked or categorized lists, particularly in a top ten or top list format

No Comments

No Known Activations