INDEX

Explanations

felt like hoursseemed like on all

New Auto-Interp

Top Features by Cosine Similarity

Configuration

Prompts (Dashboard)

10,000 prompts, 128 tokens each

Dataset (Dashboard)

lmsys/lmsys-chat-1m

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

peed

-0.09

Sta

-0.09

abit

-0.09

ullo

-0.09

-0.08

odate

-0.08

urb

-0.08

mitter

-0.08

apol

-0.08

escription

-0.08

POSITIVE LOGITS

 ages

0.14

 Ages

0.13

 hours

0.12

AGES

0.12

 longer

0.12

ilyn

0.12

lif

0.12

-et

0.12

 Longer

0.12

 seemed

0.12

Activations Density 0.019%