INDEX

Explanations

outlining steps and topics

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

 details

-0.13

 detail

-0.11

 specifics

-0.11

 detal

-0.11

Details

-0.10

 particulars

-0.10

 detalles

-0.10

_details

-0.10

ogan

-0.10

 Details

-0.10

POSITIVE LOGITS

 major

0.15

 steps

0.14

key

0.13

sal

0.13

 entire

0.12

 contents

0.11

 Entire

0.11

steps

0.11

 types

0.11

 main

0.11

Activations Density 0.099%