Pith Number

pith:CPWBRHAT

pith:2026:CPWBRHATZL3TN6PA6BA2UNH2FD

not attested not anchored not stored refs resolved

ShadowMerge: A Novel Poisoning Attack on Graph-Based Agent Memory via Relation-Channel Conflicts

Lingyun Peng, Shuyu Li, Tiantian Ji, Xinran Liu, Yang Luo, Yong Liu, Zifeng Kang

ShadowMerge poisons graph-based agent memory by injecting relations that share the same query-activated anchor and channel as legitimate evidence.

arxiv:2605.09033 v3 · 2026-05-09 · cs.CR · cs.AI

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{CPWBRHATZL3TN6PA6BA2UNH2FD}

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

SHADOWMERGE achieves 93.8% average attack success rate, improving the best baseline by 50.3 absolute points, while having negligible impact on unrelated benign tasks.

C2weakest assumption

The graph-memory system will extract, merge into the target anchor neighborhood, and retrieve the poisoned relation for the victim query when it shares the same query-activated anchor and canonicalized relation channel as benign evidence via the AIR pipeline.

C3one line summary

ShadowMerge exploits relation-channel conflicts to poison graph-based agent memory, achieving 93.8% average attack success rate on Mem0 and real-world datasets while bypassing existing defenses.

References

57 extracted · 57 resolved · 14 Pith anchors

[1] ReAct: Synergizing Reasoning and Acting in Language Models 2023 · arXiv:2210.03629

[2] Toolformer: Language Models Can Teach Themselves to Use Tools 2023 · arXiv:2302.04761

[3] Re- flexion: Language agents with verbal reinforcement learning 2023

[4] Voyager: An Open-Ended Embodied Agent with Large Language Models 2023 · arXiv:2305.16291

[5] Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology , year= 2023 · doi:10.1145/3586183.3606763

Receipt and verification

First computed	2026-05-20T00:00:41.867279Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

13ec189c13caf736f9e0f041aa34fa28e16fe313a39df369e6f1e165f678f80e

Aliases

arxiv: 2605.09033 · arxiv_version: 2605.09033v3 · doi: 10.48550/arxiv.2605.09033 · pith_short_12: CPWBRHATZL3T · pith_short_16: CPWBRHATZL3TN6PA · pith_short_8: CPWBRHAT

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/CPWBRHATZL3TN6PA6BA2UNH2FD \
  | 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: 13ec189c13caf736f9e0f041aa34fa28e16fe313a39df369e6f1e165f678f80e

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "8a981f444072b094ad1688b6e0bcbe4c6bfae8da37ba84d24ae62a67f8434b2a",
    "cross_cats_sorted": [
      "cs.AI"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CR",
    "submitted_at": "2026-05-09T16:16:41Z",
    "title_canon_sha256": "99cafecba82d31d45b38b7753fc5ed4b54dc3c30a522023f1bd61277cf64408c"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.09033",
    "kind": "arxiv",
    "version": 3
  }
}