SAE MODEL #2

FEATURE 1311

/ Molecular Mechanisms and Biological Processes

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 #6

The positive examples consistently describe research focused on molecular mechanisms, biological processes, and cellular functions. Specific themes include protein localization and function (actin-related proteins, fructose 1,6-bisphosphate aldolase), metabolic pathways (adipose conversion, glycolysis), disease mechanisms (bronchiolitis obliterans, cardiovascular calcification), and developmental processes (prenatal exposure and body composition). The language is technical and often involves experimental methods and data analysis. The negative examples, conversely, deal with topics like neuroendocrine neoplasms, antibody localization, and materials science, which lack these core features. The key discriminator is the focus on specific molecular entities and their roles in biological systems.

STATISTICS & DISTRIBUTION

Density

0.00900

Peak Act

3.75

0.0 Max

Global Context

TOP ACTIVATING CONTEXTS

DOC #917 ANALYZE

ACT: 3.7544

The actin-related protein hArp8 accumulates on the mitotic chromosomes and functions in chromosome alignment. The actin family consists of conventional actin and various actin-related proteins (Arps). Some of these Arps are localized in the nucleus, and a fraction of each of these nuclear Arps is functionally involved in chromatin remodeling and h…

DOC #694 ANALYZE

ACT: 2.5751

Bronchiolitis obliterans in workers exposed to flavoring chemicals. Medical and environmental surveys at microwave popcorn plants and flavoring production plants have revealed a risk for bronchiolitis obliterans in workers exposed to flavoring chemicals. Workers in other food industries may also be at risk. This review summarizes the available inf…

DOC #749 ANALYZE

ACT: 2.4677

Associations of prenatal exposure to Ramadan with small stature and thinness in adulthood: results from a large Indonesian population-based study. A growing body of evidence suggests that maternal diet during pregnancy can lead to permanent alterations to the physiology of the fetus. It is unknown whether intermittent maternal fasting during Ramad…

DOC #581 ANALYZE

ACT: 2.4257

Adipose conversion of 3T3-L1 cells in a serum-free culture system depends on epidermal growth factor, insulin-like growth factor I, corticosterone, and cyclic AMP. A culture system for 3T3-L1 preadipocytes based on a serum-free chemically defined medium containing fetuin, transferrin, and pantothenate is described. In this system, adipose conversi…

DOC #760 ANALYZE

ACT: 2.0516

Identification and characterization of fructose 1,6-bisphosphate aldolase genes in Arabidopsis reveal a gene family with diverse responses to abiotic stresses. Fructose 1,6-biphosphate aldolase (FBA) is a key enzyme in plants, which is involved not only in glycolysis and gluconeogenesis in the cytoplasm, but also in the Calvin cycle in plastids. R…

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).

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 1311

Semantic Interpretation

Statistics Explained

Global Context