pith. sign in
Pith Number

pith:77ND325O

pith:2026:77ND325OGD6ZCNV464EIHG64HJ
not attested not anchored not stored refs resolved

Does RAG Know When Retrieval Is Wrong? Diagnosing Context Compliance under Knowledge Conflict

Huan Xu, Pin Qian, Shuhuai Lin, Sipeng Zhang, Su Wang, Xinpeng Wei, Yihang Chen

Context-Driven Decomposition diagnoses when RAG follows conflicting retrieved context over its own knowledge.

arxiv:2605.14473 v1 · 2026-05-14 · cs.CL · cs.AI

Add to your LaTeX paper
\usepackage{pith}
\pithnumber{77ND325OGD6ZCNV464EIHG64HJ}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp
2 Internet Archive
3 Author claim open · sign in to claim
4 Citations open
5 Replications open
Portable graph bundle live · download bundle · merged state
The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

CDD exposes three patterns: context compliance is measurable in adversarial settings (Standard RAG reaches 15.0% accuracy on TruthfulQA misconception injection), adversarial accuracy gains transfer across model families but rationale-answer causal coupling does not (CDD reaches 64.1% mistake-injection causal sensitivity on Gemini-2.5-Flash while Claude variants fall in [-3%, +7%]), and explicit conflict decomposition improves robustness under temporal drift (71.3%) and noisy distractors (69.9%) on Epi-Scale.

C2weakest assumption

That the belief-decomposition probe at inference time accurately isolates the causal contribution of retrieved context to the final answer without the intervention itself altering the model's reasoning process or introducing new artifacts.

C3one line summary

Context-Driven Decomposition (CDD) measures context compliance in RAG under knowledge conflicts and improves accuracy on adversarial benchmarks like TruthfulQA misconception injection and Epi-Scale tests across models.

References

19 extracted · 19 resolved · 2 Pith anchors

[1] Alchourr´on, Peter G¨ardenfors, and David Makinson 1985
[2] Self- RAG: Learning to retrieve, generate, and critique through self-reflection 2024
[3] Rich knowledge sources bring complex knowledge conflicts: Recalibrating models to reflect conflicting evidence 2022
[4] Dense passage retrieval for open-domain question answering 2020
[5] Measuring Faithfulness in Chain-of-Thought Reasoning 2023 · arXiv:2307.13702
Receipt and verification
First computed 2026-05-17T23:39:06.636181Z
Builder pith-number-builder-2026-05-17-v1
Signature Pith Ed25519 (pith-v1-2026-05) · public key
Schema pith-number/v1.0

Canonical hash

ffda3debae30fd9136bcf708839bdc3a57a67a6b425ab93b4fe277051de3c3d8

Aliases

arxiv: 2605.14473 · arxiv_version: 2605.14473v1 · doi: 10.48550/arxiv.2605.14473 · pith_short_12: 77ND325OGD6Z · pith_short_16: 77ND325OGD6ZCNV4 · pith_short_8: 77ND325O
Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/77ND325OGD6ZCNV464EIHG64HJ \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: ffda3debae30fd9136bcf708839bdc3a57a67a6b425ab93b4fe277051de3c3d8
Canonical record JSON
{
  "metadata": {
    "abstract_canon_sha256": "f138f6f9a4eba5851f243efa10d55fa3aa4529f52b58a3d9414ee2cf1929cbf5",
    "cross_cats_sorted": [
      "cs.AI"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2026-05-14T07:14:19Z",
    "title_canon_sha256": "6c9c81e9d35063e4c00451fd0995231e84743f587ecef3eb40743ad8e0c02b20"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.14473",
    "kind": "arxiv",
    "version": 1
  }
}