INDEX

Explanations

SLATE, The challenge, Q:, Evaluation, #pragma

The neuron strongly activates on two‐letter all‐caps tokens (e.g., “SL”) often found in legal or corporate names.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

nový

-1.30

 their

-1.27

 edificio

-1.21

ﻷ

-1.19

 atau

-1.18

 templo

-1.17

だったが

-1.17

шивка

-1.13

 last

-1.10

šení

-1.10

POSITIVE LOGITS

={()=>

1.21

味付け

1.17

></

1.17

/***

1.16

 trattano

1.12

 then

1.10

 sarili

1.09

 Spani

1.07

 Technische

1.06

 eccell

1.03

Activations Density 0.000%