INDEX

Explanations

explain it to me

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

aison

-0.10

_traits

-0.09

ald

-0.09

templ

-0.09

 panc

-0.09

 Tide

-0.09

ctp

-0.09

 Panc

-0.08

PIX

-0.08

oulos

-0.08

POSITIVE LOGITS

 Carn

0.15

 speaker

0.14

 Austin

0.13

 truth

0.13

 Hosp

0.13

 Witt

0.12

Austin

0.12

 Davidson

0.12

 Verification

0.12

 intention

0.11

Activations Density 0.041%