INDEX

Explanations

but rewarding

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

edin

-0.09

oin

-0.09

itta

-0.09

 à¸ģà¸£à¸ģ

-0.09

pest

-0.08

pig

-0.08

ONUS

-0.08

»

-0.08

igen

-0.08

POSITIVE LOGITS

 rewards

0.25

 reward

0.23

 Reward

0.20

 rewarding

0.20

 rewarded

0.20

reward

0.20

 Rewards

0.19

Reward

0.18

.reward

0.14

 ultimately

0.14

Activations Density 0.039%