{"slug":"multi-agent-systems","label":"Multi-Agent Systems","item_count":12,"day_count":9,"source_count":4,"first_seen":"2026-11-02T15:36:22+00:00","last_updated":"2026-11-18T18:05:47+00:00","generated_at":"2026-11-20T00:13:34.686460+00:00","sources":["arxiv_agent_systems_research","arxiv_cs_ai","hackernews_ai","infoq_ai_ml"],"days":[{"date":"2026-11-02","items":[{"title":"An Axiomatic Framework for Reinforcement Learning in Open Multi-Agent Systems","url":"http://arxiv.org/abs/2611.03301v1","source":"arxiv_agent_systems_research","type":"paper","summary_1line":"Open multi-agent systems, where only a subset of agents are controllable, are central to real-world domains such as smart grids, yet lack a principled foundation for reinforcement learning algorithm design. We study t...","why_it_matters":"Matches feed focus: agent.","sid":"bce25dc1c825a3e7","published":"2026-11-02T15:36:22+00:00","editor_note":"Theory work on RL when only some agents are controllable; least practical item in the cluster."}]},{"date":"2026-11-04","items":[{"title":"Design Tenets for Agent Identity Management Systems","url":"https://www.nist.gov/blogs/cybersecurity-insights/design-tenets-agent-identity-management-systems","source":"hackernews_ai","type":"news","summary_1line":"Design Tenets for Agent Identity Management Systems","why_it_matters":"Matches feed focus: agent.","sid":"7e75b9009b3b0ccb","published":"2026-11-04T02:42:44+00:00","editor_note":"Only standards-body source: NIST design tenets for agent identity management."},{"title":"ServerBench: How can your coding agents design specialized LLM serving systems?","url":"http://arxiv.org/abs/2611.06841v1","source":"arxiv_cs_ai","type":"paper","summary_1line":"Large language model (LLM) coding agents develop software through experimentation, offering an opportunity to design the serving systems supporting their own operation. Effective serving requires coordinating implemen...","why_it_matters":"Matches feed focus: agentic.","sid":"37a5207d8920956c","published":"2026-11-04T17:59:07+00:00","editor_note":"Benchmark for coding agents designing LLM serving systems; adjacent, not multi-agent."}]},{"date":"2026-11-06","items":[{"title":"Coordinated Mixed-Precision KV Cache Quantization for Efficient LLM-based Multi-agent Systems","url":"http://arxiv.org/abs/2611.10096v1","source":"arxiv_agent_systems_research","type":"paper","summary_1line":"LLM-based multi-agent systems solve complex tasks through collaboration, but their growing, simultaneously resident key-value (KV) caches create a memory bottleneck. Applying single-model quantization policies indepen...","why_it_matters":"Matches feed focus: agent, eval.","sid":"d48d8aea5d45b475","published":"2026-11-06T17:07:35+00:00","editor_note":"Names KV-cache memory of co-resident agents as a bottleneck and quantizes them jointly."}]},{"date":"2026-11-09","items":[{"title":"WorkflowArena: Benchmarking Executable Workflow Orchestration in LLM-Based Multi-Agent Systems","url":"http://arxiv.org/abs/2611.12856v1","source":"arxiv_agent_systems_research","type":"paper","summary_1line":"Large language model agents are increasingly deployed as multi-agent systems rather than as a standalone model operating within a single context. Recent agent harnesses allow the model to instantiate worker agents and...","why_it_matters":"Matches feed focus: agent, harness, evaluation.","sid":"b3a45cba6445f63d","published":"2026-11-09T18:45:28+00:00","editor_note":"Benchmark for harnesses that spawn worker agents to run executable workflows."}]},{"date":"2026-11-10","items":[{"title":"On Private Knowledge Retrieval and Private Inference for secure multi-agent systems","url":"http://arxiv.org/abs/2611.13696v1","source":"arxiv_agent_systems_research","type":"paper","summary_1line":"Multi-agent systems provide a convenient framework for the massive deployment of agents based on Large Language Models (LLMs). It provides a sound framework where agents can interact with existing software to perform...","why_it_matters":"Matches feed focus: agentic, eval.","sid":"a34f1fafb710e628","published":"2026-11-10T07:43:02+00:00","editor_note":"Adds private retrieval and private inference to the multi-agent security discussion."}]},{"date":"2026-11-11","items":[{"title":"CoLFA: Collaboration-Guided Lightweight Failure Attribution for LLM-Based Multi-Agent Systems","url":"http://arxiv.org/abs/2611.14980v1","source":"arxiv_agent_systems_research","type":"paper","summary_1line":"Failure attribution in multi-agent systems requires tracing error formation to identify the decisive error step and the responsible agent. However, agents' division of work and information exchange lead to Interleaved...","why_it_matters":"Matches feed focus: agent.","sid":"ae75fdfb8dec6766","published":"2026-11-11T03:38:54+00:00","editor_note":"Lightweight method for attributing a failure to the step and agent that caused it."}]},{"date":"2026-11-13","items":[{"title":"Show HN: AgentLeak – open-source privacy testing for multi-agent LLM systems","url":"https://github.com/Privatris/AgentLeak","source":"hackernews_ai","type":"news","summary_1line":"Show HN: AgentLeak – open-source privacy testing for multi-agent LLM systems","why_it_matters":"Matches feed focus: agent.","sid":"3debe09eabc6be1b","published":"2026-11-13T01:36:36+00:00","editor_note":"First runnable tool in the thread: open-source privacy testing for multi-agent systems."},{"title":"Securing Multi-Agent Systems by Design","url":"http://arxiv.org/abs/2611.17883v1","source":"arxiv_agent_systems_research","type":"paper","summary_1line":"Multi-agent systems solve complex tasks by decomposing them into subtasks, delegating the subtasks to agents, and aggregating the agents' responses using an aggregation rule. Malicious agents can manipulate the final solution by providing malicious responses to their assigned sub","why_it_matters":"Matches feed focus: agent, evaluation.","sid":"d761218419b3b591","published":"2026-11-13T08:22:06+00:00","editor_note":"Defends aggregation of subtask answers against malicious agents."}]},{"date":"2026-11-17","items":[{"title":"Presentation: Scaling LLM-Based Ranking Systems for Latency-Critical Search & Recommendation Workloads","url":"https://www.infoq.com/presentations/scale-llm-ranking-systems-linkedin/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=AI%2C+ML+%26+Data+Engineering","source":"infoq_ai_ml","type":"news","summary_1line":"Sundara Raman Ramachandran shares how LinkedIn productionizes LLM ranking under strict P99 latency SLAs. He discusses distilling a 7B model down to 400M parameters, compressing job contexts into single-token embedding...","sid":"3449739a151bbd6a","published":"2026-11-17T11:00:00+00:00","editor_note":"Production case study on latency and distillation; tangential to multi-agent work."},{"title":"TRACE-MAS: Diagnosing and Repairing Collaboration Failures in Multi-Agent Systems","url":"http://arxiv.org/abs/2611.23362v1","source":"arxiv_agent_systems_research","type":"paper","summary_1line":"LLM-based multi-agent systems (MAS) rely on not only individual agent behaviors but also the interactions that connect agents, including message exchanges and task dependencies. Existing MAS safeguards mainly perform...","why_it_matters":"Matches feed focus: agent.","sid":"31fa9cd8fb58d57c","published":"2026-11-17T16:49:33+00:00","editor_note":"Moves past detection: targets interaction-level failures and proposes repairs."}]},{"date":"2026-11-18","items":[{"title":"Kaer: Security-Aware Replanning for Multi-Agent Systems","url":"http://arxiv.org/abs/2611.24839v1","source":"arxiv_agent_systems_research","type":"paper","summary_1line":"One popular architecture for multi-agent systems (MAS) relies on an orchestrator that generates an initial plan, coordinates its execution, and dynamically revises the plan based on agent feedback when tasks cannot be...","why_it_matters":"Matches feed focus: agentic, eval.","sid":"dfde2d652caf6073","published":"2026-11-18T18:05:47+00:00","editor_note":"Brings security checks into the orchestrator's replanning loop."}]}],"editorial":{"tldr":"This is a loose research cluster, not one event: early November papers covered RL theory, orchestration benchmarks, and KV-cache memory for co-resident agents. From Nov 4 the work tilted toward securing multi-agent systems and attributing their failures, from NIST identity tenets to CoLFA and AgentLeak.","stale":false,"whats_new":"Two new papers push from diagnosis to repair: TRACE-MAS (Nov 17) diagnoses and repairs collaboration failures between agents, and Kaer (Nov 18) adds security-aware replanning to orchestrators that revise plans mid-run.","why_it_matters":"The open problems in multi-agent stacks are now securing, debugging, and paying for them: agent identity, data leakage, which agent caused a failure, how to repair it, and KV-cache memory per resident agent.","take_for_builders":"Log which agent produced each message and plan revision, and scope each agent's credentials, so you can attribute failures and leaks; treat the repair and replanning papers as research until code is released and replicated.","beats":[{"kicker":"ORCHESTRATION","tone":"launch","headline":"RL theory for open systems and WorkflowArena, a benchmark for worker-spawning workflows","summary":"One theory paper on partially controllable agents and one executable orchestration benchmark anchor the coordination side of the thread.","sids":["bce25dc1c825a3e7","b3a45cba6445f63d"]},{"kicker":"TRUST","tone":"turn","headline":"Security and privacy become the dominant thread, from NIST guidance to an open-source leak tester","summary":"NIST identity tenets, private retrieval, secure aggregation against malicious agents, and AgentLeak all target what one agent can see, claim, or leak.","sids":["7e75b9009b3b0ccb","a34f1fafb710e628","3debe09eabc6be1b","d761218419b3b591"]},{"kicker":"COST","tone":"rising","headline":"Memory and serving cost: coordinated KV-cache quantization, agent-designed serving, LinkedIn's distillation","summary":"Only the KV-cache paper is multi-agent-specific; ServerBench and the LinkedIn talk are adjacent serving work the clustering pulled in.","sids":["37a5207d8920956c","d48d8aea5d45b475","3449739a151bbd6a"]},{"kicker":"NOW","tone":"now","headline":"From attributing failures to repairing them: CoLFA, TRACE-MAS, and Kaer","summary":"CoLFA names the faulty step and agent; TRACE-MAS repairs collaboration failures; Kaer replans with security constraints when tasks fail.","sids":["ae75fdfb8dec6766","31fa9cd8fb58d57c","dfde2d652caf6073"]}],"open_questions":["Does any open-source agent framework adopt per-agent identity along the lines of NIST's design tenets?","Does AgentLeak gain users or published results against real frameworks beyond its Show HN launch?","Does WorkflowArena get adopted as a shared orchestration benchmark by other papers?","Does coordinated KV-cache quantization ship in a serving engine that hosts several agents per GPU?","Do TRACE-MAS or Kaer release code that works with an existing orchestration framework?"],"generated_at":"2026-11-20T00:12:49Z"}}