INDEX

Explanations

implement interface

np_acts-logits-general · gemini-2.5-flash-lite

This neuron detects structural markers of question-and-answer formats—such as question numbers, option letters in parentheses, and “Q:”/“A:” labels.

oai_token-act-pair · o4-mini Triggered by @jyhe0408

structural cues of test or Q&A formatting—such as multiple‑choice option markers, Q/A labels, enumerated items, and blank/underscore fill-ins.

oai_token-act-pair · gpt-5 Triggered by @jyhe0408

formatting elements in structured documents like multiple-choice questions, code, and mathematical notation, particularly parentheses and delimiters used in lists or answer choices.

oai_token-act-pair · claude-4-5-sonnet Triggered by @jyhe0408

New Auto-Interp

Configuration

google/gemma-scope-27b-pt-res/layer_10/width_131k

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

樁

-1.65

 rése

-1.64

ڤ

-1.58

 palab

-1.57

 fiables

-1.55

 utveckling

-1.52

 évé

-1.52

 börjar

-1.50

為に

-1.50

 volledig

-1.49

POSITIVE LOGITS

 berühmten

1.61

 Trotzdem

1.48

jenige

1.46

 Bewußt

1.45

 gelieb

1.45

再度

1.44

「

1.43

 Erziehung

1.42

 berühm

1.40

 Romains