WFGY/ProblemMap/GlobalFixMap/RAG/context_drift.md

Context Drift in RAG — 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 answers alternate or degrade as dialogs get longer, even though the retriever continues to surface the right snippets.
This page stabilizes λ (semantic convergence) and prevents entropy creep in retrieval-augmented pipelines.

---

Open these first

• Visual map: RAG Architecture & Recovery
• Drift diagnostics: Entropy Collapse
• Stability checks: Logic Collapse
• Retrieval schema: Retrieval Traceability
• Payload contract: Data Contracts
• Long context limits: Memory Long Context

---

Core acceptance

• ΔS(question, retrieved) ≤ 0.45 across full chain
• λ stays convergent across 3 paraphrases and 2 seeds
• Coverage ≥ 0.70 for target section, even after N steps
• E_resonance stable on long dialog windows

---

Typical symptoms → exact fix

Symptom	Likely cause	Open this
Same question asked twice, different answers	λ drift with long chain	Entropy Collapse, Logic Collapse
Correct snippets retrieved, answer drops citation	payload contract erosion	Data Contracts, Retrieval Traceability
Paraphrase of query yields different grounding	unstable λ_observe	Retrieval Playbook
Long dialog overwrites memory	buffer collapse	Memory Long Context

---

Fix in 60 seconds

1. Three-paraphrase probe

   • Ask the same question three ways.
   • If λ flips between paraphrases, lock snippet schema and apply BBAM variance clamp.

2. ΔS check over chain

   • Log ΔS(question,retrieved) across 5–10 dialog turns.
   • If ΔS rises over time, re-segment and enforce citation-first prompting.

3. Apply module

   • Drift → BBMC + Retrieval Traceability
   • Collapse mid-chain → BBCR bridge + Logic Collapse
   • Unstable variance → BBAM clamp + Entropy Collapse

---

Copy-paste probe prompt

I uploaded TXT OS and the WFGY Problem Map.

My issue:
- same query gives different answers in long dialog
- traces: ΔS(question,retrieved)=..., λ states across 3 paraphrases

Tell me:
1) where context drift occurs,
2) the exact WFGY page to open,
3) the minimal fix to enforce convergence,
4) a reproducible test over 5 turns.

---

🔗 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	Canonical framework entry point	View
Problem Map	Diagnostic map and navigation hub	View
Tension Universe Experiments	MVP experiment field	View
Recognition	Where WFGY is referenced or adopted	View
AI Guide	Anti-hallucination reading protocol for tools	View

If this repository helps, starring it improves discovery for other builders.