Yang Liu

Weifeng Yuan, Wenbo Guo, Feng Dong et al.

LLM agents acquire new capabilities by downloading skills from open registries. Instead of browsing these catalogs manually, developers typically ask the agent to recommend and install a skill. This convenience hides a risk: agents frequently invent names for skills that exist in no registry. We term this flaw skill name hallucination. A fake name may seem harmless, but it opens the door to supply-chain attacks. Because registries rarely verify publishers, an adversary can prompt the agent, collect the fake names it returns, pre-register malicious skills under them, and wait for a victim to install the payload. We conducted the first large-scale measurement of skill name hallucination, evaluating 15,000 prompts across 12 configurations (4 standalone LLMs and 8 agents). We conservatively counted a name as hallucinated only if it was missing from all live registries and GitHub. The results reveal a systemic vulnerability: every configuration hallucinates. Rates average 36.0% for standalone LLMs and 36.9% for agents, rising to 43.1% on real-world developer questions. In total, the systems generated 5,669 distinct hallucinated names. Crucially, these names are not random noise. Agents repeat the same fake names across prompts and models, giving attackers highly reliable targets to hijack. Finally, we tested four model-level defenses and found a severe conflict between security and usability. The strongest, retrieval grounding, cut the hallucination rate from 40.8% to 3.2% but crippled usefulness: even the best-defended system recommended the correct skill only about one in six times. Skill name hallucination is thus a highly exploitable vulnerability requiring minimal attacker effort. Fixing it cannot rely on prompt engineering or model tuning alone. It demands ecosystem-wide structural changes: registry-level name reservations and verified recommendation pipelines.

Liangdong Lu, Ruipan Yang, Yang Liu et al.

We introduce a Jordan-canonical-form framework for constructing $q$-ary quantum stabilizer codes from arbitrary classical linear codes over $\F_{q^2}$. The framework does not require the classical linear code $\mathcal{C}$ to satisfy the dual-containing condition (i.e., self-orthogonality). Given a classical code $\mathcal{C}=[n,k,d]_{q^2}$ with parity-check matrix $H$, we measure the obstruction to Hermitian self-orthogonality by the rank $r=(n-k)-\dim_{\F_{q^2}}(\mathcal{C}^{\perp_h}\cap \mathcal{C})$. The ingredient code $\mathcal{C}$ is $r$-nearly dual containing, or, equivalently, $\mathcal{C}^{\perp_h}$ is $r$-nearly self-orthogonal, by which we mean that $r=\Rank(HH^{\dagger})=\dim_{\F_{q^2}}(\mathcal{C}^{\perp_h})-\dim_{\F_{q^2}}(\mathcal{C}^{\perp_h}\cap \mathcal{C})$. By systematically reducing the rank of the Hermitian inner-product matrix $A=HH^{\dagger}$ through rank-one perturbations along the Jordan basis $W=P^{-1}$ of the decomposition $A=PJ_AP^{-1}$, we construct an explicit Hermitian self-orthogonal code $\mathcal{C}_{\mathrm{so}}=[n+r,n-k]_{q^2}$. A sufficient distance-preservation criterion guarantees that the resulting $q$-ary quantum code has parameters $[[n+r,2k-n+r,\geq d]]_q$. Applying this construction to classical codes produces several record quantum codes that improve or supplement the best-known parameters in Grassl's tables.

Jing Liu, Kun Yang, Yan Wang et al.

Agentic AI systems are reshaping communications and networking by deploying autonomous intelligent agents capable of collaborative learning while maintaining data privacy at network edges. Within distributed network environments, Multimodal Large Language Models (MLLMs) serve as cognitive engines for edge devices, yet federated fine-tuning faces substantial challenges in balancing global knowledge aggregation with local adaptation under heterogeneous network conditions. Conventional federated protocols typically rely on uniform parameter aggregation, which conflates domain-invariant features with client-specific nuances, thereby resulting in suboptimal personalization and excessive communication overhead. To address these challenges, we propose PFAdapter, a communication-efficient framework introducing hierarchical LoRA decomposition to explicitly separate adapter parameters into global-shared and local-private components. Query and key projections are assigned to global synchronization for capturing universal multimodal semantics across the network, while value and output projections remain localized for edge-specific adaptation. Additionally, orthogonality regularization based on the Frobenius norm enforces strict separation between these components, preventing redundant feature learning. Selective aggregation protocols synchronize only global-shared components across the federated network, preserving local expertise and reducing communication costs by nearly 50%. Extensive experiments on VQA-RAD, SLAKE, Hateful Memes, and CrisisMMD datasets demonstrate that PFAdapter consistently outperforms state-of-the-art baselines, achieving accuracy improvements ranging from 2.4% to 4.8% across diverse edge intelligence tasks. Consequently, our framework establishes an efficient solution for agentic AI deployment in resource-constrained communication networks.