1. The Problem: Structural Hallucination
When you ask a Large Language Model to generate a structured report — eight sections, in a specific order, with specific headers — the model will often comply. Mostly. It might generate seven sections. Or merge two into one. Or write the correct headers in the wrong language. Or silently drop a header and run two sections together in a single block of prose.
We call this Structural Hallucination: the tendency of a probabilistic model to produce output that is semantically correct but structurally malformed. Unlike factual hallucination (where the model invents false information), structural hallucination is harder to detect and more destructive to production systems — it breaks downstream parsers silently.
In the SAGE Oracle engine, a "clumped" or missing section means a user receives a reading that is incomplete or unrenderable. The application fails not because the AI was wrong, but because the AI was structurally imprecise.
Standard solutions — better prompting, few-shot examples, stricter temperature settings — reduce the failure rate but cannot eliminate it. A probabilistic model will always have a non-zero failure probability per generation. At scale, across 10 languages and thousands of users, that probability becomes a certainty.
2. The Solution: Post-Processing as a Structural Firewall
Rather than engineering around the LLM (using better prompts) or through it (using multiple API calls to self-correct), the SAGE engine engineers after it. We treat the LLM as an unreliable narrator and pass every response through a deterministic enforcement pipeline before it ever touches the user interface.
The architecture rests on a fundamental insight: even when an LLM fails to write the correct structural marker, the semantic content of each section is still present in the output. A reading that is missing "Section 3: Planetary Influences" still contains the prose about planetary influences — it's just untagged.
If we can identify semantic content reliably, we can synthetically reconstruct the missing structural markers from the prose itself — without re-querying the LLM.
3. The Six-Stage Pipeline
The Structural Integrity Enforcement Pipeline (SIEP) operates in six distinct stages, each targeting a specific failure mode:
Ingestion & Prompt Echo Purge
The raw LLM response is received and immediately scanned for "prompt echoes" — instances where the model has repeated back its own instructions. These are stripped via pattern matching before any structural analysis begins.
Doctrine Firewall & Header Shielding
A locale-aware registry of canonical section headers is applied. Any header that has been malformed, partially translated, or omitted is replaced with the exact canonical string for the target locale. Figure names are also normalized against a hard-coded vocabulary dictionary to prevent LLM-invented terminology from corrupting the doctrinal purity of the reading.
Anchor-Based Synthetic Header Recovery
This is the core innovation. If a section header is entirely absent, the engine scans the surrounding prose for Semantic Anchors — domain-specific keywords unique to each section (e.g., specific life-domain terminology for Section 4). When an anchor is detected, the corresponding canonical header is synthetically injected at the correct position, recovering the section without any LLM retry.
Hyper-Isolated Translation Shielding
For multilingual output, content is wrapped in opaque positional markers ([PART-N]) before being passed to the translation model. The markers are explicitly excluded from translation, preventing the translation LLM from corrupting the established segment structure. A multi-variant regex ([\[【「]?PART\s*-\s*(\d+)) handles the marker variations that different LLMs naturally produce.
Token Desquashing
Cross-lingual generation causes technical tokens to "squash" — fusing with adjacent words. A fuzzy pattern normalizer identifies and repairs these fusions (e.g., H1-Tariq → H1: Tariq), maintaining machine-readable data integrity. A hallucinated placeholder repair system also fixes malformed tokens like {{RF_Tariq using role-deduplication logic.
Automated Multi-Metric QA Validation
A final validation loop runs 12 independent checks: header order verification, section count, bullet counts per section, script purity ratio (to detect language leaks), Mojibake detection, prose repetition similarity, and technical token leakage checks. Only responses passing all checks are delivered to the user interface.
4. Results
After deploying the SIEP across all 10 supported languages (English, Hindi, Arabic, Chinese, Japanese, French, German, Spanish, Portuguese, Russian), we observed the following outcomes:
100% schema compliance — Every delivered reading follows the exact 8-section architecture.
Zero language leaks — No English prose detected in non-English readings post-pipeline.
Reduced latency — By eliminating LLM retry loops, average response time decreased significantly. The structural repair is computationally trivial compared to a re-generation API call.
5. The Broader Application
The SIEP was designed for the SAGE Geomancy engine, but the architecture is domain-agnostic. Any application that requires structured, deterministic output from a non-deterministic generative model faces the same problem. This includes medical report generation, legal document drafting, multilingual customer support systems, and structured data extraction pipelines.
The key principle — treat the LLM as an unreliable narrator and enforce structure deterministically downstream — is applicable wherever the cost of structural failure is high and re-querying the model is expensive or undesirable.
Jolly, A. (2026). Solving Structural Hallucination in Generative AI: The Structural Integrity Enforcement Pipeline (SIEP). Zenodo. https://doi.org/10.5281/zenodo.19945234
6. Conclusion
Structural Hallucination is an underexplored failure mode in production LLM systems. While the research community focuses on factual accuracy, real-world deployments frequently fail due to structural imprecision. The SIEP demonstrates that a deterministic post-processing firewall, grounded in semantic anchors and canonical registries, can eliminate structural hallucination without sacrificing generation quality or requiring model fine-tuning.
The engineering behind SAGE is not just about ancient wisdom — it is about proving that AI can be made reliable enough for precision applications where "mostly correct" is never acceptable.
Experience the Oracle
The engineering described in this paper powers every reading on SAGE. Try it for free and see the difference structural precision makes.
✧ Get 11 Free Energy Tokens ✧