INDEX

Explanations

idea of

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

ASK

-0.09

indsight

-0.09

åĩ¡

-0.09

 hete

-0.08

 Hindered

-0.08

åĲįçĦ¡ãģĹãģķãĤĵ

-0.08

_FP

-0.08

_reviews

-0.08

cae

-0.08

POSITIVE LOGITS

 mention

0.27

 reference

0.25

 mentions

0.22

mention

0.21

 references

0.20

 notion

0.18

 Mention

0.18

åħ³äºİ

0.18

 idea

0.17

 mentioned

0.17

Activations Density 0.173%