INDEX

Explanations

Avoiding complexity in writing

np_max-act · gemini-2.0-flash

This neuron responds to terms that criticize or qualify writing—words like “convoluted,” “phrasing,” “repetition,” “unjustified,” “complexity,” “warnings,” and “generated,” i.e. negative-style or meta‐instruction keywords.

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

늘

-0.06

_TS

-0.06

Ice

-0.06

cool

-0.06

Mp

-0.06

は

-0.06

 Euros

-0.06

;*/↵

-0.06

 ClassNotFoundException

-0.06

 کاملا

-0.06

POSITIVE LOGITS

 medida

0.07

ipzig

0.07

ッチ

0.07

 theology

0.06

野

0.06

प

0.06

actors

0.06

ophilia

0.06

educt

0.06

adolu

0.06

Activations Density 0.013%

Avoiding complexity in writing

This neuron responds to terms that criticize or qualify writing—words like “convoluted,” “phrasing,” “repetition,” “unjustified,” “complexity,” “warnings,” and “generated,” i.e. negative-style or meta‐instruction keywords.

No Comments

No Known Activations

Avoiding complexity in writing

This neuron responds to terms that criticize or qualify writing—words like “convoluted,” “phrasing,” “repetition,” “unjustified,” “complexity,” “warnings,” and “generated,” i.e. negative-style or meta‐instruction keywords.

No Comments

No Known Activations