INDEX

Explanations

gibberish text

np_max-act · gemini-2.0-flash

This neuron activates on words in the assistant’s standard “gibberish detection” phrase (e.g. “random,” “sequence,” “letters,” “and”) that say the user’s input is a meaningless string of characters.

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

Negative Logits

apy

-0.07

anes

-0.06

┃

-0.06

shade

-0.06

rière

-0.06

َي

-0.06

 WEIGHT

-0.06

Malloc

-0.06

_fonts

-0.06

ToDate

-0.06

POSITIVE LOGITS

 senators

0.07

oğunluk

0.07

 enrollment

0.07

 silicon

0.06

ฤษภาคม

0.06

 percentages

0.06

skb

0.06

 aktiv

0.06

χω

0.06

weeks

0.06

Activations Density 0.021%

gibberish text

This neuron activates on words in the assistant’s standard “gibberish detection” phrase (e.g. “random,” “sequence,” “letters,” “and”) that say the user’s input is a meaningless string of characters.

No Known Activations

gibberish text

This neuron activates on words in the assistant’s standard “gibberish detection” phrase (e.g. “random,” “sequence,” “letters,” “and”) that say the user’s input is a meaningless string of characters.

No Known Activations