INDEX

Explanations

who

np_max-act · gemini-2.0-flash

The neuron never fires (always zero activation), so it does not detect any feature.

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

expressions of apology and confusion in conversational contexts.

oai_token-act-pair · gpt-4o-mini Triggered by @xinyanhu8

the transition between scenes or dialog turns, particularly in text involving interactive conversation or role-playing elements.

oai_token-act-pair · gpt-4o 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

 dụng

-0.07

uation

-0.06

에서

-0.06

 또는

-0.06

اجه

-0.06

 transmission

-0.06

accumulator

-0.06

 indication

-0.06

 zwischen

-0.06

TM

-0.06

POSITIVE LOGITS

�

0.07

Ин

0.07

 klíč

0.07

¬

0.07

 mantener

0.07

losing

0.06

 возмож

0.06

 Güney

0.06

目を

0.06

Activations Density 0.031%

who

The neuron never fires (always zero activation), so it does not detect any feature.

expressions of apology and confusion in conversational contexts.

the transition between scenes or dialog turns, particularly in text involving interactive conversation or role-playing elements.

No Comments

No Known Activations