INDEX

Explanations

with/that/which

np_max-act · gemini-2.0-flash

The neuron fires on the part of the prompt that instructs to answer using only a single letter (e.g. “with only single letter like A, B or C”), i.e. it detects the “single‐letter answer” directive.

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

Not in Any Lists

No Comments

Negative Logits

 broth

-0.07

 procur

-0.06

 fails

-0.06

 dragon

-0.06

:Add

-0.06

urray

-0.06

almö

-0.06

 próxima

-0.06

 occurrence

-0.06

crc

-0.06

POSITIVE LOGITS

%%%%

0.07

 trí

0.07

	super

0.07

[class

0.06

كوم

0.06

чика

0.06

 connector

0.06

 زیر

0.06

uated

0.06

	index

0.06

Activations Density 0.001%

with/that/which

The neuron fires on the part of the prompt that instructs to answer using only a single letter (e.g. “with only single letter like A, B or C”), i.e. it detects the “single‐letter answer” directive.

No Comments

No Known Activations

with/that/which

The neuron fires on the part of the prompt that instructs to answer using only a single letter (e.g. “with only single letter like A, B or C”), i.e. it detects the “single‐letter answer” directive.

No Comments

No Known Activations