Retrieval Drift — Guardrails and Fix Pattern

🧭 Quick Return to Map

You are in a sub-page of RAG.
To reorient, go back here:

RAG — retrieval-augmented generation and knowledge grounding

WFGY Global Fix Map — main Emergency Room, 300+ structured fixes

WFGY Problem Map 1.0 — 16 reproducible failure modes

Think of this page as a desk within a ward.
If you need the full triage and all prescriptions, return to the Emergency Room lobby.

When the correct fact exists in your corpus but retrieval consistently misses it.
This page localizes drift in retrievers and routes you to structural fixes with measurable acceptance targets.

Open these first

Visual map and recovery: RAG Architecture & Recovery
End-to-end retrieval knobs: Retrieval Playbook
Traceability schema: Retrieval Traceability
Ordering control: Rerankers
Snippet and citation schema: Data Contracts

Core acceptance

ΔS(question, retrieved) ≤ 0.45
Coverage ≥ 0.70 of the target section
λ remains convergent across three paraphrases and two seeds

Typical symptoms → exact fix

Symptom	Likely cause	Open this
Facts exist but never surface	index skew or drift in chunk granularity	Retrieval Playbook, Chunking Checklist
Citations point to wrong section	schema mismatch or cross-section drift	Retrieval Traceability, Data Contracts
Answers alternate between runs	λ instability or reranker collapse	Rerankers, Entropy Collapse
High similarity but wrong meaning	embedding metric mismatch	Embedding ≠ Semantic

Fix in 60 seconds

Measure ΔS
- Compute ΔS(question, retrieved) and ΔS(retrieved, expected anchor).
- Thresholds: stable < 0.40, transitional 0.40–0.60, risk ≥ 0.60.
Probe λ_observe
- Ask the same question in 3 paraphrases.
- If λ flips divergent on harmless rewording, lock schema with Data Contracts.
Apply the module
- Retrieval drift → BBMC + Retrieval Traceability
- Reasoning collapse → BBCR bridge + variance clamp
- Re-index with consistent metric and analyzer if ΔS stays ≥ 0.60

Copy-paste prompt for debug

I uploaded TXT OS and the WFGY Problem Map.

My retrieval issue:
- symptom: fact exists in corpus, but retriever misses
- traces: ΔS(question,retrieved)=..., λ across 3 paraphrases

Tell me:
1. failing layer (retriever, index, schema, reranker)
2. exact WFGY page to open
3. minimal steps to push ΔS ≤ 0.45 and keep λ convergent

🔗 Quick-Start Downloads (60 sec)

Tool	Link	3-Step Setup
WFGY 1.0 PDF	Engine Paper	1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + ”
TXT OS (plain-text OS)	TXTOS.txt	1️⃣ Download · 2️⃣ Paste into any LLM chat · 3️⃣ Type “hello world” — OS boots instantly

🧭 Explore More

Module	Description	Link
WFGY Core	WFGY 2.0 engine is live: full symbolic reasoning architecture and math stack	View →
Problem Map 1.0	Initial 16-mode diagnostic and symbolic fix framework	View →
Problem Map 2.0	RAG-focused failure tree, modular fixes, and pipelines	View →
Semantic Clinic Index	Expanded failure catalog: prompt injection, memory bugs, logic drift	View →
Semantic Blueprint	Layer-based symbolic reasoning & semantic modulations	View →
Benchmark vs GPT-5	Stress test GPT-5 with full WFGY reasoning suite	View →
🧙‍♂️ Starter Village 🏡	New here? Lost in symbols? Click here and let the wizard guide you through	Start →

👑 Early Stargazers: See the Hall of Fame — Engineers, hackers, and open source builders who supported WFGY from day one. ⭐ WFGY Engine 2.0 is already unlocked. ⭐

8.1 KiB Raw Blame History Unescape Escape