INDEX

Explanations

setting goals and rewards

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

 astr

-0.10

 Spending

-0.09

pond

-0.09

 spending

-0.09

æĮĩå¯¼

-0.09

èĥĨ

-0.09

atten

-0.09

apore

-0.08

NU

-0.08

ieves

-0.08

POSITIVE LOGITS

 reward

0.24

 rewards

0.21

 Reward

0.19

 Rewards

0.18

Reward

0.17

reward

0.17

 rewarded

0.16

 rewarding

0.15

_reward

0.13

 Find

0.13

Activations Density 0.054%