INDEX

Explanations

subject of research or debate

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

Negative Logits

obel

-0.09

 Quest

-0.08

canf

-0.08

odash

-0.08

 footprint

-0.08

Chu

-0.08

ainless

-0.08

WebHost

-0.08

 watt

-0.08

ogram

-0.08

POSITIVE LOGITS

 focus

0.17

 interest

0.16

 discussion

0.15

 study

0.15

 concern

0.14

 attention

0.13

interest

0.12

çĦ¦

0.12

 contention

0.12

Interest

0.11

Activations Density 0.045%

No Known Activations