INDEX

Explanations

chemical intermediate/precursor

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

ayar

-0.10

hecy

-0.09

adir

-0.09

ilik

-0.09

uos

-0.08

asher

-0.08

pler

-0.08

adh

-0.08

richt

-0.08

awah

-0.08

POSITIVE LOGITS

 starting

0.26

raw

0.22

starting

0.20

 Starting

0.17

Raw

0.16

Starting

0.16

raw

0.15

 base

0.15

key

0.13

Raw

0.13

Activations Density 0.016%