INDEX

Explanations

expressions that solicit feedback or input from the audience

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ãģ¾ãģ¾

-0.16

rema

-0.15

 Becker

-0.14

strup

-0.14

 DÃ¼n

-0.14

got

-0.14

sten

-0.13

tvrt

-0.13

.rpm

-0.13

ingle

-0.13

POSITIVE LOGITS

 know

0.29

 hear

0.23

 knows

0.20

 known

0.17

 aware

0.16

know

0.16

 hearing

0.16

 savoir

0.16

 what

0.16

 Know

0.15

Activations Density 0.020%

expressions that solicit feedback or input from the audience

No Comments

No Known Activations