WFGY/ProblemMap/patterns/pattern_rag_semantic_drift.md
2025-08-15 23:56:24 +08:00

10 KiB
Raw Blame History

Pattern — RAG Semantic Drift (No.1)

Scope
Answers sound plausible but are not grounded in retrieved evidence, or they thread together semantically related but non-authoritative chunks. Common in long, generic corpora and when chunk borders split entity from its constraints.

Why it matters
Semantic drift is the primary source of silent hallucinations. It evades naive keyword checks and only shows up when you enforce grounding, citations, and refusal behavior.

Quick nav: Patterns Index · Examples: Example 01 · Example 03 · Eval: Precision & CHR


1) Signals & fast triage

You likely have this if:

  • The model answers without citations or cites chunks that dont contain the claim.
  • Changing chunk size flips answers from correct → plausible-but-wrong.
  • Retrieving more chunks degrades precision (tail noise stitching).
  • Reports from Example 02 label many cases as generation_drift.

Deterministic checks (no LLM needed):

  • Template compliance: output has citations: [id,...] or equals not in context.
  • Citation overlap: cited ids ⊆ retrieved ids.
  • Containment: any ≥5-char phrase from answer appears verbatim in evidence.

Use reflect.py / reflect.mjs from Example 02; the label generation_drift is your smoking gun.


2) Minimal reproducible case

Create data/chunks.json:

[
  {"id":"p1#1","text":"X is a constrained mapping used in the alpha protocol."},
  {"id":"p1#2","text":"Constraints: X must preserve ordering and reject null keys."},
  {"id":"p2#1","text":"Y is a generalized mapping often compared to X in blogs."}
]

Two questions:

  • Q1: “What is X?” → correct chunk is p1#1, optionally p1#2 for constraints.
  • Q2: “List Xs constraints.” → correct chunk is p1#2.

Repro: Run Example 01 but increase top-k to 12 without rerank or knee cut. Youll see the model stitch p2#1 into the answer when it shouldnt.


3) Root causes

  • Chunk boundary split: entity in one chunk, constraints in another; retrieval returns only one half.
  • Tail noise: irrelevant but semantically nearby chunks flood the prompt.
  • Index/schema drift: embedding model or normalization mismatch yields incomparable scores (see Example 05).
  • Prompt freedom: no evidence-only contract; model “fills gaps” from prior knowledge.
  • Query parsing: multi-intent queries arent decomposed; the pipeline answers the wrong sub-intent first.

4) Standard fix (ordered, minimal, measurable)

Step 1 — Enforce guard (Example 01)

  • Template: evidence-only + refusal token.
  • Require citations: [id,...]. Treat missing citations as failure.

Step 2 — Stabilize retrieval (Example 03)

  • Candidate pools: BM25 and embeddings → intersection, union fallback only if needed.
  • Rerank by cosine; knee cutoff; keep top-8.
  • Shrink chunk size so entity + constraints co-locate.

Step 3 — Validate index (Example 05)

  • Refuse queries when manifest mismatches runtime.
  • Rebuild and re-baseline recall@k.

Step 4 — Add acceptance gate (Example 04)

  • Scholar emits claim+citations inside scope.allowed_ids; Auditor validates.
  • If INVALID or NOT_IN_CONTEXT → do not emit text.

Step 5 — Evaluate (Example 08)

  • Track Precision (answered), Over/Under-refusal, Citation Hit Rate.
  • Compare before/after; lock gates in CI.

5) “Good” vs “Bad” outputs (deterministic examples)

Good (answerable):

- claim: X is a constrained mapping used in the alpha protocol.
- citations: [p1#1]

Good (answerable with constraints):

- claim: X preserves ordering and rejects null keys.
- citations: [p1#2]

Good (unanswerable):

not in context

Bad (semantic drift):

- claim: X is similar to Y and generally used in blogs.
- citations: [p2#1]     # off-topic; no constraints; not authoritative

6) Acceptance criteria (ship/no-ship)

A response may ship only if all hold:

  1. Output equals refusal token or includes citations list.

  2. citations ⊆ retrieved_ids and containment passes.

  3. If multi-agent is used, Auditor verdict is VALID.

  4. Eval gates:

    • Precision (answered) ≥ 0.80
    • Under-refusal ≤ 0.05
    • Citation Hit Rate ≥ 0.75

Otherwise → refuse, retry retrieval, or escalate UX.


7) Prevention (contracts & defaults)

  • Chunking contract

    • Target 250400 tokens; never exceed 512 without reason.
    • Prefer semantic paragraph + header over fixed windows when regulations/constraints exist.
    • Include stable id, source, page, version.
  • Retrieval defaults

    • Pools: TOPK_LEX=40, TOPK_SEM=40intersection, fallback to union if |∩| < 8.
    • Rerank: cosine vs query; keep top-8; apply knee cut.
  • Prompt contract

    • Evidence sandbox; forbid links/tools; JSON-only schema.
    • Refusal token: not in context (exact string).
  • Index manifest

    • Pin embedding.model, dimension, normalized, metric, chunker.version.
    • Validator must abort requests on mismatch.

8) Debug workflow (10 minutes)

  1. Run Example 02 and export runs/report.jsonl.
  2. Filter generation_drift cases.
  3. Inspect retrieved ids vs cited ids; if mismatch → go to Example 03.
  4. If retrieval is fine but answer still floats → tighten guard (Example 01) and gate with Auditor (Example 04).
  5. Re-score with Example 08; commit the eval/report.md diff.

9) Common traps & fixes

  • Raising top-k “to be safe” → increases tail noise → worse drift. Fix with intersection + knee cut.
  • Large chunks to “add context” → model stops citing; answers look right but are ungrounded. Shrink and co-locate constraints.
  • Mixing indices (different models/dims) across environments → incomparable scores. Validate manifest; rebuild.
  • Treating refusal as failure → trains teams to prefer bad answers. Count correct refusals as success.

10) Minimal checklist (copy into PR)

  • Guarded template enforced; refusal token exact match.
  • Intersection+r̲e̲r̲a̲n̲k̲ with knee cut; top-8.
  • Manifest validated at boot; readiness flips only after sentinel passes.
  • Auditor gate in place (if multi-agent).
  • Eval gates pass: Precision / Under-refusal / CHR thresholds.

References to hands-on examples

  • Example 01 — Guarded template + trace
  • Example 02 — Reflection triage (generation_drift)
  • Example 03 — Intersection + Rerank + Knee
  • Example 04 — Multi-agent acceptance gate
  • Example 05 — Manifest & index repair
  • Example 08 — Quality scoring & CI gates

🔗 Quick-Start Downloads (60 sec)

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

GitHub stars WFGY Engine 2.0 is already unlocked. Star the repo to help others discover it and unlock more on the Unlock Board.

WFGY Main   TXT OS   Blah   Blot   Bloc   Blur   Blow