INDEX

Explanations

failure or error conditions

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

Mismatch

-0.12

Incorrect

-0.11

rias

-0.10

enco

-0.10

zy

-0.10

NotFound

-0.09

 Mayo

-0.09

ISMATCH

-0.09

lim

-0.09

ollen

-0.09

POSITIVE LOGITS

 failed

0.17

 Something

0.16

 unable

0.15

 couldn

0.15

 something

0.14

Something

0.14

 difficulty

0.14

 fails

0.13

failed

0.13

 failure

0.13

Activations Density 0.040%