INDEX

Explanations

science experiments and bombs

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

 Dummy

-0.11

 imagin

-0.11

 barrels

-0.10

 metabol

-0.10

 PLUGIN

-0.09

 haystack

-0.09

imagin

-0.09

 Cousins

-0.09

ampler

-0.09

 mocked

-0.09

POSITIVE LOGITS

 experiment

0.17

 science

0.15

 experiments

0.14

 Science

0.13

 household

0.13

experiment

0.12

 conduct

0.12

 Experiment

0.12

Science

0.12

 Newton

0.12

Activations Density 0.130%