INDEX

Explanations

key ideas or takeaways

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

elsen

-0.09

 recycled

-0.09

Neu

-0.09

dep

-0.08

oba

-0.08

rega

-0.08

orig

-0.08

_GF

-0.08

à¹Īà¸§à¸¡

-0.08

trie

-0.08

POSITIVE LOGITS

 take

0.40

 Take

0.39

 takeaway

0.39

Take

0.36

take

0.36

 TAKE

0.33

.take

0.28

_take

0.27

.Take

0.23

key

0.22

Activations Density 0.119%