INDEX

Explanations

I did, is being, went have

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

 bitten

-0.10

LOPT

-0.09

IIIK

-0.09

ÐĴÐ¡

-0.09

TSR

-0.08

itled

-0.08

 amen

-0.08

 migrationBuilder

-0.08

ÐĴÑĤ

-0.08

alace

-0.08

POSITIVE LOGITS

".

0.11

 indeed

0.10

0.09

ahren

0.09

 unspecified

0.09

 Indeed

0.09

 apparently

0.08

 certain

0.08

vl

0.08

Activations Density 0.093%