INDEX

Explanations

The neuron fires on filesystem‐style path fragments (e.g. strings containing “/” and directory or file names).

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

domains for HIV inhibitors

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

technical syntax or notation in code, file paths, and scientific measurements.

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

New Auto-Interp

Configuration

google/gemma-scope-27b-pt-res/layer_22/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

 also

-1.05

 here

-1.00

You

-0.94

”

-0.92

ten

-0.89

nand

-0.86

you

-0.85

 Column

-0.84

ko

-0.83

POSITIVE LOGITS

写真は

1.04

 périph

0.97

ICAGO

0.95

 despertar

0.91

昨日の

0.91

 olvido

0.91

hộp

0.91

 gröss

0.90

昨日は

0.89

前回の

0.89

Activations Density 0.009%

The neuron fires on filesystem‐style path fragments (e.g. strings containing “/” and directory or file names).

domains for HIV inhibitors

technical syntax or notation in code, file paths, and scientific measurements.

No Comments

No Known Activations

The neuron fires on filesystem‐style path fragments (e.g. strings containing “/” and directory or file names).

domains for HIV inhibitors

technical syntax or notation in code, file paths, and scientific measurements.

No Comments

No Known Activations