INDEX

Explanations

references to the concept of "hijacking."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 31-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L31R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.31.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

issan

-0.16

ifiant

-0.16

aires

-0.16

 Baton

-0.16

 Prem

-0.15

esan

-0.15

wand

-0.15

aston

-0.14

auga

-0.14

elow

-0.14

POSITIVE LOGITS

acking

0.43

acked

0.42

ackers

0.37

acker

0.37

inks

0.34

inx

0.34

ack

0.31

hij

0.29

acks

0.26

0.25

Activations Density 0.010%

references to the concept of "hijacking."

No Comments

No Known Activations