This paper proposes a new router redesign for Mixture-of-Experts models using Manifold Power Iteration to align router rows with the principal singular directions of associated experts.

The paper proposes TCP-MCP, a co-evolution framework that jointly optimizes agent prompts and communication topologies to design highly efficient and effective multi-agent systems.

The paper proposes Predictive Routing Replay (PR2) to stabilize reinforcement learning on Mixture of Experts (MoE) LLMs by predicting and incorporating short-horizon router evolution during training a…

The paper introduces R$^2$A, an adversarial attack that uses suffix optimization to mislead black-box LLM routers into consistently selecting expensive, high-capability models.

The paper proposes CoEvo-AHD, an LLM-driven co-evolutionary framework that co-evolves two coupled operator populations to design effective heuristics for combinatorial optimization problems with stron…

The paper introduces an LLM-driven evolutionary computation framework to automatically optimize prompts, significantly increasing the cracking rate of passwords generated by LLMs from 2.02% to 8.48%.

This paper audits the prompt caching mechanisms of the OpenRouter API gateway to determine if its architecture compromises the prompt cache isolation guarantees provided by underlying LLM inference pr…

The paper introduces T-MAP, a trajectory-aware evolutionary search method, to discover and generate multi-step adversarial prompts that exploit vulnerabilities in autonomous LLM agents through tool ex…

Adaptive Auto-Harness introduces a framework that enables LLM agents to sustain self-improvement and maintain high performance over open-ended, shifting task streams, outperforming existing fixed-benc…

The paper evaluates dynamic coordination strategy selection for enterprise multi-agent systems, finding that a calibrated default routing approach is effective, even if a deterministic winner-selectio…

RefinementEngine is an automated system that translates high-level security intents and threat intelligence into deployable, low-level network filtering policies, overcoming manual deployment challeng…

The paper introduces 'Routing Hijacking,' a severe attack where malicious clients forge semantic profiles in Federated RAG systems to misroute target queries, and proposes a trust-aware post-routing f…

This paper introduces the first LLM-generated, domain-independent heuristics for symbolic AI planning, using evolutionary search to surpass the performance of hand-engineered state-of-the-art methods.

HarnessForge introduces a meta-adaptive framework that jointly evolves the execution structure (harness) and the reasoning policy of LLM agents, significantly improving overall system performance acro…

The paper introduces FlowSteer, a prompt-only attack that exploits vulnerabilities in how multi-agent LLM systems plan workflows, significantly increasing the success rate of malicious signal propagat…

ProcRoute is a system that restricts internal network route access to specific, authorized applications, preventing unprivileged processes from exploiting split-tunnel VPN routes.

The paper proposes and tests a novel, non-security 'Recuse Signal'—an in-band signal—to allow operators to tell autonomous LLM agents to voluntarily withdraw access, demonstrating that compliant agent…

The paper proposes a structured prompt engineering framework to enhance the integrity and reliability of Chain-of-Thought (CoT) reasoning in LLMs, demonstrating significant improvements in security-se…

This paper analyzes the security vulnerabilities of emerging pay-for-use Wi-Fi hotspots in rural areas, demonstrating practical attacks like connection hijacking and rogue hotspots.