{ "schema_version": "0.6.0", "paper_id": "taaf-trace-abstraction-analysis-framework", "page_url": "https://naser.github.io/research-publications/papers/taaf-trace-abstraction-analysis-framework/", "title": "TAAF: A Trace Abstraction and Analysis Framework Synergizing Knowledge Graphs and LLMs", "title_variants": [], "authors": [ "Alireza Ezaz", "Ghazal Khodabandeh", "Majid Babaei", "Naser Ezzati-Jivan" ], "author_details": [ { "name": "Alireza Ezaz", "orcid": null, "profile_url": "https://dblp.org/pid/374/6221.html" }, { "name": "Ghazal Khodabandeh", "orcid": null, "profile_url": "https://dblp.org/pid/374/6434.html" }, { "name": "Majid Babaei", "orcid": null, "profile_url": "https://dblp.org/pid/19/10401.html" }, { "name": "Naser Ezzati-Jivan", "orcid": "https://orcid.org/0000-0003-1435-6297", "profile_url": "https://naser.github.io/" } ], "publication": { "year": 2026, "venue": "arXiv / CoRR; arXiv page reports acceptance to ICSE 2026", "type": "preprint with publication signal", "publication_date": "2026", "online_date": null, "print_date": null, "volume": "abs/2601.02632", "issue": null, "pages": null, "article_number": null, "publisher": null, "issn": [], "isbn": [], "crossref_type": null }, "publication_type": "preprint with publication signal", "status": "public_arxiv_preprint_with_publication_signal", "canonical_source_url": "https://doi.org/10.1145/3744916.3787832", "source_record_id": "taaf-a-trace-abstraction-and-analysis-framework-synergizing-knowledge-graphs-and-llms-00e7e01d64", "identifiers": { "doi": "10.1145/3744916.3787832", "arxiv": "2601.02632", "arxiv_doi": "10.48550/arXiv.2601.02632" }, "abstract": null, "abstract_source": "official arXiv:2601.02632 abstract", "abstract_available": false, "scholar_eligibility": { "eligible": false, "basis": "not-eligible", "note": "The page is a discovery record; it does not claim Google Scholar article-host eligibility." }, "description": "TAAF combines time-indexed knowledge graphs with large language models to turn large execution traces into queryable, actionable explanations.", "evidence_level": "full-text-reviewed", "evidence": { "source_basis": "full-text-reviewed", "coverage": "material paper sections", "summary_origin": "AI-assisted catalog editorial summary", "review_status": "catalog-reviewed; paper-author approval pending", "verified_on": "2026-08-09", "sources": [ { "note": "official arXiv abstract, metadata, and PDF page 2601.02632" }, { "note": "arXiv page comment containing the ICSE 2026 acceptance and ACM DOI" }, { "note": "local PDF hash verified in the working catalog" } ] }, "summary": { "core_contribution": "TAAF combines time-indexed knowledge graphs with large language models to turn large execution traces into queryable, actionable explanations.", "problem": "Kernel and large-application traces are too large, multidimensional, and temporally structured for direct manual inspection or raw LLM ingestion. Existing stateful tools support predefined queries, but flexible questions still require domain-specific scripts and expert knowledge.", "method": "TAAF implements a three-layer trace-to-answer pipeline: LTTng events are transformed into a time-indexed Trace Compass-style State System; a natural-language query selects a compact temporal knowledge graph with typed entities, weighted relations, and time scope; a schema-conditioned LLM answers from the graph. The pipeline is evaluated with TraceQA-100, a benchmark of expert-authored questions over real kernel traces.", "findings": "TraceQA-100 contains 100 questions in explanatory, multiple-choice, and true/false formats, split across single- and multi-hop reasoning. Across the core GPT-4.1 nano/GPT-4o/o4-mini grid and 1/10/100-second windows, graph grounding improves weighted accuracy by a reported mean 21.5 percentage points and up to 31.17 points; o4-mini with TAAF reaches 95.5% at 1 second and 90.17% at 100 seconds. Supplying the graph schema adds 8.1 points in the focused GPT-4o test.", "limitations": "The benchmark is hand-crafted from SciMark 2.0 traces under Linux and uses a three-level score, author labeling, three samples per configuration, API models, and a raw-State-System baseline only. Long windows, explanatory multi-hop arithmetic, global queries, model drift, and other kernels/workloads remain difficult. The methods section accounts for 5,400 Phase-1 plus 2,100 Phase-2 outputs (7,500), while the conclusion says 7,800; preserve this as an unresolved paper-internal inconsistency.", "future_work": "Broaden question generation and annotation, use finer numeric-tolerance metrics and independent raters, support temporal/production-scale traces, hierarchical or retrieval-based prompting, autonomous-agent integration, and optimize KG construction with partitioning, streaming, batching, and early exit." }, "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" ], "versions": [ { "id": "publication-doi-listed-on-arxiv-page", "label": "Publication DOI listed on arXiv page", "relation": "version-of-record", "title": "TAAF: A Trace Abstraction and Analysis Framework Synergizing Knowledge Graphs and LLMs", "url": "https://doi.org/10.1145/3744916.3787832", "pdf_url": null, "status": "publication_signal", "canonical_for_citation": true }, { "id": "public-preprint", "label": "Public preprint", "relation": "preprint", "title": "TAAF: A Trace Abstraction and Analysis Framework Synergizing Knowledge Graphs and LLMs", "url": "https://arxiv.org/abs/2601.02632", "pdf_url": "https://arxiv.org/pdf/2601.02632", "status": "public_full_text", "canonical_for_citation": false } ], "access": { "status": "public_arxiv_preprint_with_publication_signal", "note": "The arXiv page reports acceptance to ICSE 2026 and lists the ACM DOI; the publication record should be rechecked when the final proceedings metadata is available.", "license": null }, "resources": { "code": null, "data": null, "slides": null, "demo": null }, "citation_guidance": { "when_to_cite": "Cite this paper when your work uses or compares the State System -> query-specific temporal knowledge graph -> LLM architecture for grounded kernel-trace question answering.", "points": [ "the State System -> query-specific temporal knowledge graph -> LLM architecture for grounded kernel-trace question answering.", "TraceQA-100, including expert-authored temporal, multi-entity, numeric, single-hop, and multi-hop questions over SciMark 2.0 kernel traces.", "the measured benefit of graph grounding, especially the reported +21.5-point mean gain and up-to +31.17-point gain over raw State-System output.", "the schema and temporal-window ablations when discussing how explicit structure and shorter windows affect LLM trace reasoning." ], "canonical_version_id": "publication-doi-listed-on-arxiv-page" }, "provenance": { "metadata_verified_on": "2026-08-09", "metadata_source": [ "official arXiv abstract, metadata, and PDF page 2601.02632", "arXiv page comment containing the ICSE 2026 acceptance and ACM DOI", "local PDF hash verified in the working catalog" ], "summary_written_by": "AI-assisted", "summary_verified_by": "full-text-grounded catalog review; author approval pending", "linked_preprint_record": null, "author_order_note": null }, "batch": { "phase": 2, "batch_label": "expanded forty-paper release", "status": "included_in_expanded_catalog", "selected_at": "2026-08-09" } }