INDEX

Explanations

brackets and symbols

np_max-act · gemini-2.0-flash

The neuron is primarily activated by numeric tokens (integers and decimal numbers).

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

structured number patterns with punctuation separators (like dates/times, counts, and quantifier-style ranges)

oai_token-act-pair · gpt-5 Triggered by @vetterc0

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

 fabric

-0.07

 melhor

-0.06

 evil

-0.06

queue

-0.06

 paddle

-0.06

орд

-0.06

 forfe

-0.06

 setStatus

-0.06

ービ

-0.06

survey

-0.06

POSITIVE LOGITS

0.07

 níž

0.06

 ratt

0.06

 ));↵↵

0.06

.signIn

0.06

AL

0.06

 Česká

0.06

 nás

0.06

ieren

0.06

	Class

0.06

Activations Density 2.207%

brackets and symbols

The neuron is primarily activated by numeric tokens (integers and decimal numbers).

structured number patterns with punctuation separators (like dates/times, counts, and quantifier-style ranges)

No Comments

No Known Activations