INDEX

Explanations

code

np_max-act · gemini-2.0-flash

This neuron primarily detects structural and formatting cues in instructions—e.g. code-block markers (backticks), JSON labels, and directive words that frame how content should be formatted.

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

 보기

-0.06

 EXPER

-0.06

rollo

-0.06

�니다

-0.06

fn

-0.06

ivering

-0.06

 trif

-0.06

dato

-0.06

ahren

-0.06

 صورت

-0.06

POSITIVE LOGITS

 기록

0.06

.feature

0.06

 tão

0.06

कन

0.06

PAR

0.06

gsl

0.06

.getSystemService

0.06

 salvar

0.06

HDATA

0.06

 встанов

0.06

Activations Density 0.049%

code

This neuron primarily detects structural and formatting cues in instructions—e.g. code-block markers (backticks), JSON labels, and directive words that frame how content should be formatted.

No Comments

No Known Activations