{
  "schema_version": "0.1.0",
  "record_type": "research-topic",
  "topic_id": "llm-assisted-analysis",
  "label": "LLM-Assisted Analysis",
  "title": "LLM-Assisted Systems Analysis Research",
  "description": "Research papers on large language models for trace analysis, observability, reasoning, and systems diagnosis.",
  "introduction": "This topic groups work that uses large language models to query, summarize, reason over, or govern system evidence. The papers distinguish the trace or structured representation provided to the model, the reasoning task, evaluation data, and the limits of language-model assistance for technical diagnosis.",
  "aliases": [
    "LLM systems analysis",
    "large language models for tracing",
    "LLM-assisted observability"
  ],
  "search_terms": [
    "LLM-assisted analysis",
    "large language models",
    "trace question answering",
    "LLM observability",
    "agentic systems"
  ],
  "related_topics": [
    "trace-analysis",
    "trace-abstraction",
    "observability",
    "knowledge-graphs",
    "llm-evaluation"
  ],
  "canonical_url": "https://threadslab.org/research-publications/topics/llm-assisted-analysis.html",
  "paper_count": 5,
  "papers": [
    {
      "paper_id": "lmat-adaptive-tracing-language-models",
      "title": "LMAT: An adaptive tracing approach based on efficient system behavior analysis using language models",
      "year": 2026,
      "authors": [
        "Kasra Darvishi",
        "Morteza Noferesti",
        "Yuvraj Sehgal",
        "Naser Ezzati-Jivan"
      ],
      "page_url": "https://threadslab.org/research-publications/papers/lmat-adaptive-tracing-language-models/",
      "canonical_source_url": "https://doi.org/10.1016/j.jss.2026.112890",
      "core_contribution": "LMAT combines multi-task language models for kernel-event and event-duration prediction with online change detection, lightweight error-vector root-cause analysis, and adaptive tracing control.",
      "tags": [
        "kernel-tracing",
        "system-tracing",
        "llm-assisted-analysis",
        "observability",
        "anomaly-detection",
        "root-cause-analysis",
        "trace-reduction"
      ],
      "keywords": [
        "adaptive tracing",
        "LTTng",
        "kernel events",
        "system-call sequences",
        "event-duration modeling",
        "language models",
        "LSTM",
        "BERT",
        "multi-task learning",
        "HDBSCAN",
        "change detection",
        "root-cause analysis",
        "Apache",
        "Sock Shop",
        "OpenTelemetry",
        "Prometheus",
        "stress-ng",
        "trace reduction",
        "trace overhead"
      ]
    },
    {
      "paper_id": "observability-runtime-governance-agentic-ai-systems",
      "title": "Observability and Runtime Governance for Agentic AI Systems",
      "year": 2026,
      "authors": [
        "Naser Ezzati-Jivan",
        "Maryam Ekhlasi"
      ],
      "page_url": "https://threadslab.org/research-publications/papers/observability-runtime-governance-agentic-ai-systems/",
      "canonical_source_url": "https://doi.org/10.1145/3803437.3804904",
      "core_contribution": "The tutorial presents an end-to-end AgentOps workflow that connects task intent and model decisions to tool calls, memory access, inter-agent communication, external side effects, and runtime-governance decisions.",
      "tags": [
        "observability",
        "llm-assisted-analysis",
        "performance-analysis"
      ],
      "keywords": [
        "agentic AI",
        "AgentOps",
        "runtime governance",
        "AI observability",
        "software agents",
        "agent tracing",
        "silent failures",
        "goal drift",
        "tool-use failures",
        "cross-layer evidence",
        "policy checks",
        "human escalation",
        "FSE 2026"
      ]
    },
    {
      "paper_id": "rethinking-performance-debugging-from-optimization-to-collaborative-reasoning",
      "title": "Rethinking Performance Debugging: From Optimization to Collaborative Reasoning",
      "year": 2026,
      "authors": [
        "Mahsa Panahandeh",
        "Naser Ezzati-Jivan",
        "Abdelwahab Hamou-Lhadj"
      ],
      "page_url": "https://threadslab.org/research-publications/papers/rethinking-performance-debugging-from-optimization-to-collaborative-reasoning/",
      "canonical_source_url": "https://doi.org/10.1145/3803437.3805578",
      "core_contribution": "The paper reframes performance debugging as collaborative reasoning over multiple evidence-grounded hypotheses rather than optimization for one supposedly best explanation.",
      "tags": [
        "performance-analysis",
        "microservices",
        "llm-assisted-analysis",
        "root-cause-analysis",
        "observability"
      ],
      "keywords": [
        "performance debugging",
        "collaborative reasoning",
        "AgentDebug",
        "Reasoning Surface",
        "hypothesis generation",
        "hypothesis critique",
        "hypothesis curation",
        "Gemini 2.5 Flash",
        "CART",
        "DeLag",
        "Train-Ticket",
        "E-Shopper",
        "fault injection",
        "RPC latency",
        "Pareto-optimal hypotheses"
      ]
    },
    {
      "paper_id": "taaf-trace-abstraction-analysis-framework",
      "title": "TAAF: A Trace Abstraction and Analysis Framework Synergizing Knowledge Graphs and LLMs",
      "year": 2026,
      "authors": [
        "Alireza Ezaz",
        "Ghazal Khodabandeh",
        "Majid Babaei",
        "Naser Ezzati-Jivan"
      ],
      "page_url": "https://threadslab.org/research-publications/papers/taaf-trace-abstraction-analysis-framework/",
      "canonical_source_url": "https://doi.org/10.1145/3744916.3787832",
      "core_contribution": "TAAF combines time-indexed knowledge graphs with large language models to turn large execution traces into queryable, actionable explanations.",
      "tags": [
        "kernel-tracing",
        "trace-analysis",
        "knowledge-graphs",
        "llm-assisted-analysis",
        "observability"
      ],
      "keywords": [
        "TAAF",
        "TraceQA-100",
        "time-indexed knowledge graph",
        "natural-language trace queries",
        "multi-hop reasoning",
        "causal reasoning"
      ]
    },
    {
      "paper_id": "toward-adaptive-tracing-language-models",
      "title": "Toward Adaptive Tracing: Efficient System Behavior Analysis using Language Models",
      "year": 2024,
      "authors": [
        "Kasra Darvishi",
        "Morteza Noferesti",
        "Naser Ezzati-Jivan"
      ],
      "page_url": "https://threadslab.org/research-publications/papers/toward-adaptive-tracing-language-models/",
      "canonical_source_url": "https://doi.org/10.1145/3639476.3639778",
      "core_contribution": "The paper proposes an adaptive tracing loop that combines language-model prediction of kernel-event sequences and event durations with change detection and root-cause analysis, so detailed tracing is activated only around significant behavior shifts.",
      "tags": [
        "kernel-tracing",
        "system-tracing",
        "llm-assisted-analysis",
        "observability",
        "anomaly-detection",
        "root-cause-analysis",
        "trace-reduction"
      ],
      "keywords": [
        "adaptive tracing",
        "LTTng",
        "kernel events",
        "system-call sequences",
        "event-duration modeling",
        "language models",
        "LSTM",
        "BERT",
        "multi-task learning",
        "change detection",
        "root-cause analysis",
        "Apache2",
        "trace reduction",
        "trace overhead"
      ]
    }
  ]
}