SAE MODEL #2

FEATURE 1302

/ Sensing and Detection Technologies

This page shows the detailed analysis of a specific feature in the Sparse Autoencoder (SAE). It includes the semantic interpretation generated by the LLM, the top activating documents that trigger this feature, and statistical metrics like density and activation distribution.

Semantic Interpretation

gemma3:12b #7

The positive examples consistently describe devices and methods for sensing, detecting, and analyzing various substances or phenomena. This includes light-emitting devices, hydrophones, microfluidic chips for chemical detection, acoustic localization techniques, biosensors for bacteria detection, nanoscale vibration sensors, and related applications. The core theme revolves around using physical or chemical principles to identify and quantify specific targets or properties. The negative examples, conversely, deal with biological and medical research topics (wortmannin inhibition, cerebellar neuron activity, rodent carcinogenesis, tumor oxygenation) that do not fit this sensing/detection paradigm.

STATISTICS & DISTRIBUTION

Density

0.01700

Peak Act

3.75

0.0 Max

Global Context

TOP ACTIVATING CONTEXTS

DOC #101 ANALYZE

ACT: 3.7531

Let it shine: a transparent and photoluminescent foldable nanocellulose/quantum dot paper. Exploration of environmentally friendly light-emitting devices with extremely low weight has been a trend in recent decades for modern digital technology. Herein, we describe a simple suction filtration method to develop a transparent and photoluminescent na…

DOC #549 ANALYZE

ACT: 2.7530

DFB fiber laser hydrophone with band-pass response. A distributed-feedback fiber laser hydrophone with band-pass response is presented. The design of the hydrophone aims to equalize static pressure and eliminate signal aliasing of high-frequency acoustic components. Theoretical analysis is presented based on electro-acoustic theory. The experiment…

DOC #455 ANALYZE

ACT: 2.6857

A fully integrated passive microfluidic Lab-on-a-Chip for real-time electrochemical detection of ammonium: Sewage applications. The present work reports on the development of a new generation of Lab-on-a-chip (LOC) to perform in-situ and real-time potentiometric measurements in flowing water. The device consisted of two differentiated parts: a pol…

DOC #829 ANALYZE

ACT: 2.5205

Note: Localization based on estimated source energy homogeneity. Acoustic signal localization is a complex problem with a wide range of industrial and academic applications. Herein, we propose a localization method based on energy attenuation and inverted source amplitude comparison (termed estimated source energy homogeneity, or ESEH). This inver…

DOC #404 ANALYZE

ACT: 2.4921

Label-free visual biosensor based on cascade amplification for the detection of Salmonella. Salmonella is a widely distributed, extremely harmful bacteria, the presence of which requires confirmation via an on-site visual biosensor. In this study, we constructed a label-free, cascade amplification visualization biosensor for the sensitive and rapi…

TOPOLOGY

ROOT OF FAMILY #4

Interactions and Mechanisms #404

CORRELATIONS

W – Weight-space · similarity between decoder vectors (features that point in similar directions in the embedding space).

D – Data / co-activation · features that tend to fire together on the same documents (co-occurrence in the dataset).

Interactions and Mechanisms 0.12
Nanomaterial Synthesis and Properties 0.06
Medical Imaging and Oncology 0.06

Ollama Model

Temp

K+

K-

System Prompt

You are a meticulous researcher investigating a specific neuron in a language model. Your task is to determine what behavior this neuron is responsible for: what concepts, topics, or linguistic features does it activate on?

INPUT DESCRIPTION: You will receive two inputs: 1) Maximum Activation Examples and 2) Zero Activation Examples.
1. You will be given several text examples that activate the neuron, along with a number indicating how strongly it was activated. This means there is some feature, concept, or pattern in this text that 'excites' this neuron.
2. You will also be given several text examples that do NOT activate the neuron. This means the feature or concept is not present in these texts.

OUTPUT DESCRIPTION: Given the inputs provided, complete the following tasks.
1. Based on the MAXIMUM ACTIVATION EXAMPLES, list potential topics, concepts, themes, and features they have in common. Be specific. You may need to look at different levels of granularity. List as many as possible. Give greater weight to concepts more prominent in higher-activation examples.
2. Based on the zero activation examples, systematically exclude any topic/concept/feature listed above that also appears in the zero activation examples.
3. Based on the two previous steps, perform a thorough analysis of which feature, concept, or topic, at which level of granularity, is likely to activate this neuron. Use Occam's razor, as long as it fits the evidence provided. Be highly rational and analytical.
4. Based on step 3, summarize this concept in 1-8 words, in the form FINAL: <explanation>. Do NOT return anything after these 1-8 words.

Respond EXCLUSIVELY with valid JSON: {'label': '...', 'description': '...'}

FEATURE 1302

Semantic Interpretation

Statistics Explained

Global Context