Section Balance checks whether the document is chunked in a maintainable way. It penalizes:Documentation Index
Fetch the complete documentation index at: https://mehen.ophi.dev/llms.txt
Use this file to discover all available pages before exploring further.
- Oversized sections at the 95th percentile.
- A high rate of very large sections.
- A high rate of tiny sections.
- Heading skips (e.g.,
H2 → H4). - Heading depth deviation from a profile-specific expectation.
Inputs
- The section tree.
- Per-section word counts from the LOC family.
- Profile-specific expected depth (default depends on doc type).
How DMI uses it
Section Balance feeds DMI via theS_norm term — “poor section balance”
reduces DMI.
References
- Miller, G. A. (1956). The magical number seven, plus or minus two: some limits on our capacity for processing information. Psychological Review 63(2): 81–97 — working-memory ceiling that motivates the oversized-section penalty. DOI.
- Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science 12(2): 257–285 — cognitive-load theory underlying the “too-many-tiny-sections vs. too-large-section” trade-off. DOI.
See also
- Section tree — input structure.
- Filler / Lazy Risk — separate “lazy sectioning” sub-score.
- Good Scaffold — orthogonal “did the scaffolding help?” axis.