INDEX

Explanations

error handling with except

The neuron strongly responds to contrastive or concessive cue words and negation (e.g. “though,” “but,” “if,” “not,” “won’t,” “stop”).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

},

0.61

}$,

0.59

ondes

0.59

bote

0.59

Примеча

0.59

}])

0.59

other

0.58

ियारों

0.57

었고

0.57

 coincided

0.55

POSITIVE LOGITS

 سعی

1.26

Try

1.24

try

1.23

 versuchen

1.23

 Trying

1.20

TRY

1.18

尝试

1.13

嘗試

1.12

try

1.09

 versucht

1.08

Activations Density 0.015%