INDEX

Explanations

numbers

np_max-act · gemini-2.0-flash

The neuron strongly activates on runs of uppercase letters (especially multi‐letter “cipher” or acronym‐style tokens).

oai_token-act-pair · o4-mini Triggered by @xinyanhu8

New Auto-Interp

Configuration

andyrdt/saes-llama-3.1-8b-instruct/resid_post_layer_11/trainer_1

Dataset (Dashboard)

Various

Features

131,072

Data Type

float32

Hook Name

blocks.11.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

یشه

-0.07

۱۵

-0.06

FIT

-0.06

 evolutionary

-0.06

jag

-0.06

４

-0.06

 ты

-0.06

 polling

-0.06

 Routing

-0.06

Iran

-0.06

POSITIVE LOGITS

 Captain

0.07

 Illustrated

0.07

Với

0.07

 accompagn

0.07

 사람

0.06

azers

0.06

 eskort

0.06

 fined

0.06

("'",

0.06

.schema

0.06

Activations Density 0.018%

numbers

The neuron strongly activates on runs of uppercase letters (especially multi‐letter “cipher” or acronym‐style tokens).

No Comments

No Known Activations