INDEX

Explanations

role indicator then content

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

ë»

-0.09

ibr

-0.09

akter

-0.08

ä¹ĥ

-0.08

::::::::::::::::

-0.08

strand

-0.08

 seating

-0.08

ãĥ³ãĤº

-0.08

itzer

-0.08

 dryer

-0.08

POSITIVE LOGITS

 Richt

0.09

 EXTRA

0.08

.Helper

0.08

ssa

0.08

è¾ĳ

0.08

eck

0.08

 ench

0.08

 Crate

0.08

cce

0.08

antino

0.08

Activations Density 0.414%