INDEX

Explanations

references to criminal history and legal offenses

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 24-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L24R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.24.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

hower

-0.17

 TORT

-0.17

weeney

-0.16

inish

-0.15

abler

-0.15

.minecraft

-0.15

bose

-0.14

ãĥ©ãĥĥãĤ¯

-0.14

perse

-0.14

esda

-0.14

POSITIVE LOGITS

 conviction

0.45

 convictions

0.43

 Conv

0.38

conv

0.37

Conv

0.35

 conv

0.32

_conv

0.29

.conv

0.28

 convicted

0.27

 convex

0.26

Activations Density 0.046%

references to criminal history and legal offenses

No Comments

No Known Activations

references to criminal history and legal offenses

No Comments

No Known Activations