INDEX

Explanations

references to news sources or news-related terminology

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 23-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L23R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.23.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

HEME

-0.16

ÐµÑĢÐ³

-0.16

erville

-0.15

ÑĩÐµÐ½

-0.15

istica

-0.15

oad

-0.15

oops

-0.14

arium

-0.14

obus

-0.14

ester

-0.14

POSITIVE LOGITS

room

0.19

 outlets

0.18

Hour

0.17

stands

0.17

 outlet

0.16

night

0.16

plex

0.16

week

0.16

osaur

0.15

om

0.15

Activations Density 0.019%

references to news sources or news-related terminology

No Comments

No Known Activations