FEATURE 1792

/ Metastatic Cancer

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 #11

The neuron strongly activates in contexts describing cancer that has spread from its original site to other parts of the body, particularly to bone, liver, muscle, and the meninges. The examples consistently involve descriptions of diagnostic procedures (CT scans, biopsies, scintigraphy), treatment options (liver transplantation, esophagectomy), and the progression of the disease. The specific types of cancer mentioned (melanoma, prostate, breast, colorectal, pancreatic, gastric) are less important than the overarching theme of cancer metastasis. The presence of terms like 'metastasis,' 'bone metastases,' 'liver metastases,' 'skeletal muscle metastasis,' and 'leptomeningeal metastasis' are key indicators. The descriptions often include details about diagnostic imaging, surgical interventions, and the monitoring of tumor markers. The negative examples focus on unrelated medical topics like cardiac surgery, lipid management, and molecular biology, which lack this metastatic cancer theme.

STATISTICS & DISTRIBUTION

Density

0.02900

Peak Act

3.71

0.0 Max

Global Context

TOP ACTIVATING CONTEXTS

DOC #594 ANALYZE

ACT: 3.7090

Bone Metabolism of the Patient with a Malignant Melanoma during the Entry Examination and the Check-up of Whole-body Bone Scintigraphy. Malignant melanoma is a malignancy located predominantly in the skin and the incidence of melanoma increases. We compared the markers of bone metabolism - osteocalcin (OC), beta-carboxyterminal cross-linked telope…

DOC #94 ANALYZE

ACT: 3.2574

CT of intracranial metastases with skull and scalp involvement. Twenty-eight persons with contiguous intracranial skull, and often extracranial metastatic disease are reported. These lesions comprised 7.6% of a series of 250 consecutive patients with intracranial metastatic disease. Only three of 28 patients had other intracranial lesions and only…

DOC #44 ANALYZE

ACT: 2.8195

[Clinical analysis of liver transplantation in treatment of liver metastatic cancer]. To evaluate the feasibility and outcome of liver transplantation in the treatment of liver metastatic cancer. Four patients with pathologically confirmed liver metastatic cancer underwent liver transplantation, including one of liver metastases from pancreatic en…

DOC #889 ANALYZE

ACT: 2.6015

[A case of skeletal muscle metastasis on the left thigh after esophagectomy for esophageal adenocarcinoma in Barrett' s esophagus]. We report the case of a 45-year-old man with advanced esophageal adenocarcinoma in Barrett's esophagus. After neoadjuvant chemotherapy using S-1, docetaxel, and cisplatin, the patient underwent thoracoscopic esophagec…

DOC #71 ANALYZE

ACT: 2.3519

Leptomeningeal metastasis with urological presentation. The incidence of clinical leptomeningeal metastases from non-neurologic solid tumour is generally increasing as a result of better survival of systemic cancer with chemotherapy. Their presentation varies according to the neurological involvement. A case of leptomeningeal metastasis from an un…

TOPOLOGY

ROOT OF FAMILY #1

Medical Case Reports & Cancer Research #2232

Medical Imaging and Oncology #374
Child of Drug Resistance Mechanisms in Cancer
#708

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).

Medical Imaging and Oncology 0.34
Drug Resistance Mechanisms in Cancer 0.14
Feature 164 0.07
Medical Case Reports & Cancer Research 0.07
Feature 1822 0.03
Feature 2555 0.03

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 1792

Semantic Interpretation

Statistics Explained

Global Context