INDEX

Explanations

discussing instructions

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

 Strange

-0.09

 Combination

-0.08

 magn

-0.08

 blur

-0.08

 options

-0.08

anson

-0.08

 logic

-0.08

oup

-0.08

 fame

-0.08

oster

-0.08

POSITIVE LOGITS

 instruction

0.29

 instructions

0.27

 Ð¸Ð½ÑģÑĤÑĢÑĥÐº

0.22

 Instructions

0.20

instructions

0.19

Instructions

0.18

_instructions

0.18

 Instruction

0.18

instruction

0.17

.instructions

0.16

Activations Density 0.096%