PRISMΔDB
SAE MODEL #2

FEATURE 2603

/ Instructional/Educational Content
Refresh Stats

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 #4

The neuron appears to activate in response to text that provides instructions, advice, or explanations related to learning, skill development, or processes. This includes topics like succeeding in online courses, understanding research methodologies, improving writing skills, or developing training programs. The presence of phrases like 'tips for,' 'practical advice,' 'process of,' 'development of,' and 'assessment of' are strong indicators. The content often involves a 'how-to' element, or a description of a procedure or methodology. The negative examples lack this instructional or explanatory tone, focusing instead on research reports, medical case studies, or guidelines.

STATISTICS & DISTRIBUTION
Statistics Explained

Density
Fraction of documents where this feature activates at least once.
Higher density = feature appears frequently across the dataset.

Peak Activation
Maximum activation value observed for this feature over all documents.

Activation Histogram
Distribution of all activation values for this feature. Each bar represents a bin (range) of values, and its height shows how many documents fall in that range.

Density
0.00200
Peak Act
3.76
0.0 Max
Global Context
TOP ACTIVATING CONTEXTS
DOC #268 ANALYZE
ACT: 3.7612
Tips for succeeding at Internet courses. Ready to go to "cyberschool"? Here's practical advice on making it work for you.
DOC #481 ANALYZE
ACT: 2.0278
Comparison of V[Combining Dot Above]O2peak Performance on a Motorized vs. a Nonmotorized Treadmill. Morgan, AL, Laurent, CM, and Fullenkamp, AM. Comparison of V[Combining Dot Above]O2peak performance on a motorized vs. a nonmotorized treadmill. J Strength Cond Res 30(7): 1898-1905, 2016-Despite growing popularity of nonmotorized treadmills (NMTs),…
DOC #333 ANALYZE
ACT: 1.5446
Maintaining motivation and health among recreational runners: Panel study of factors associated with self-rated performance outcomes at competitions. To investigate health-related factors associated with self-rated race performance outcomes among recreational long-distance runners. Panel study. Data were collected from runners one month before and…
DOC #429 ANALYZE
ACT: 1.4564
Development of an introductory course in child protection. The maltreatment of children is a significant public health and social problem. Healthcare professionals have a crucial role to play working with other agencies to protect children from abuse and neglect. The need for training, support and clinical supervision in this work has been identif…
DOC #369 ANALYZE
ACT: 1.4544
Writing to get published. Writing for publication is an important way to contribute to the knowledge base for professional nurses. This paper reviews the process of writing a professional paper from identification of a topic to the time the paper appears in print. Web-based resources to facilitate the writing process are highlighted.
DOC #883 ANALYZE
ACT: 1.4373
Assessment of competency in clinical measurement: comparison of two forms of sequential test and sensitivity of test error rates to parameter choice. To assess clinical measurement competency by two sequential test formulations [resetting sequential probability ratio test (R-SPRT) and learning curve cumulative summation (LC-CUSUM)]. Numerical simu…
TOPOLOGY
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).