INDEX

Explanations

episodes, EP.1, epistemological

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

cet

-0.10

 Neutral

-0.10

on

-0.09

ar

-0.09

ugar

-0.09

jam

-0.09

liv

-0.09

lib

-0.09

EXP

-0.09

lec

-0.09

POSITIVE LOGITS

ep

0.17

,ep

0.15

Ep

0.14

(ep

0.14

ymous

0.12

stein

0.11

RLF

0.11

Î³Î³ÎµÎ»

0.11

worth

0.11

erves

0.10

Activations Density 0.035%