INDEX

Explanations

phrases that emphasize a high level of quality or recognition

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

æĹ¢

-0.17

plain

-0.16

 really

-0.15

 plain

-0.15

imax

-0.15

inc

-0.15

iful

-0.14

uste

-0.14

 easy

-0.14

/how

-0.14

POSITIVE LOGITS

 regarded

0.23

 caffe

0.18

-reg

0.18

 dziew

0.17

 styl

0.17

 combust

0.16

irth

0.16

-special

0.16

 regiment

0.15

 regard

0.15

Activations Density 0.016%

phrases that emphasize a high level of quality or recognition

No Comments

No Known Activations