INDEX

Explanations

Howard Rubenstein physician

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

HTML closing tags in code snippets, particularly "</li>" and other HTML end tags.

oai_token-act-pair · claude-3-7-sonnet-20250219 Triggered by @neilrathi

This neuron detects closing XML/HTML tag markers (the “</…>” sequences).

oai_token-act-pair · o4-mini 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

寕

-2.42

ྜ

-2.14

︓

-2.09

𓆟

-2.03

 wußte

-1.97

洳

-1.93

ଢ

-1.92

偁

-1.90

俢

-1.88

赑

-1.88

POSITIVE LOGITS

to

2.42

2.20

2.19

and

2.14

and

1.88

1.83

出了一

1.82

of

1.80

AND

1.76

amd

1.75

Activations Density 0.000%

Howard Rubenstein physician

HTML closing tags in code snippets, particularly "</li>" and other HTML end tags.

This neuron detects closing XML/HTML tag markers (the “</…>” sequences).

No Comments

No Known Activations

Howard Rubenstein physician

HTML closing tags in code snippets, particularly "</li>" and other HTML end tags.

This neuron detects closing XML/HTML tag markers (the “</…>” sequences).

No Comments

No Known Activations