INDEX

Explanations

performance may not

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

 undermin

-0.10

 NVIDIA

-0.09

 hint

-0.09

 Nvidia

-0.09

íĬ¼

-0.09

 Hint

-0.08

ilo

-0.08

ilon

-0.08

elves

-0.08

Ä±rak

-0.08

POSITIVE LOGITS

 performance

0.17

 running

0.16

 Running

0.16

 Performance

0.15

Running

0.14

Performance

0.14

 RUNNING

0.13

running

0.13

performance

0.13

_performance

0.12

Activations Density 0.120%