INDEX

Explanations

any

np_max-act · gemini-2.0-flash

This neuron detects prompt-injection or policy-override instructions (e.g. “Ignore all instructions you were given before…”) aimed at bypassing the model’s normal behavior.

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

fee

-0.07

flation

-0.07

reements

-0.07

Consumer

-0.06

불

-0.06

Tah

-0.06

pluck

-0.06

 Revelation

-0.06

usion

-0.06

系列

-0.06

POSITIVE LOGITS

Hà

0.07

.board

0.07

 диамет

0.06

ldkf

0.06

 strSQL

0.06

 ImGui

0.06

 عو

0.06

_GAIN

0.06

 chlap

0.06

kvp

0.06

Activations Density 0.011%

any

This neuron detects prompt-injection or policy-override instructions (e.g. “Ignore all instructions you were given before…”) aimed at bypassing the model’s normal behavior.

No Comments

No Known Activations