LongDS-Bench: On the Failure of Long-Horizon Agentic Data Analysis

• URL: http://arxiv.org/abs/2605.30434v1
• Date: Thu, 28 May 2026 18:00:20 GMT
• Title: LongDS-Bench: On the Failure of Long-Horizon Agentic Data Analysis
• Abstract summary: LongDS is a benchmark for long-horizon, multi-turn data analysis.<n>Agents must maintain, update, restore, and compose evolving analytical states.<n>LongDS comprises 68 tasks constructed from real-world Kaggle notebooks.
• Score: 29.918718173329037
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Real-world data analysis is inherently iterative, yet existing benchmarks mostly evaluate isolated or short interactive tasks, leaving agents' ability to track evolving analytical context over long horizons untested. We introduce LongDS, a benchmark for long-horizon, multi-turn data analysis where agents must maintain, update, restore, and compose evolving analytical states. LongDS comprises 68 tasks constructed from real-world Kaggle notebooks, spanning 2,225 turns across six domains including Geoscience, Business, and Education. Tasks are designed around state-evolution patterns (e.g., counterfactual perturbation, rollback, multi-state composition), with an average dependency span of 11.3 turns. Evaluating five state-of-the-art models, we find that the best model reaches only 48.45% average accuracy, performance drops nearly 47 points from early to late turns, and long-horizon errors account for 52%--69% of failures. Further analysis shows that additional agent steps do not necessarily improve performance, suggesting that the key bottleneck is maintaining a correct analytical state rather than increasing interaction budget. We release LongDS to support research on reliable long-horizon agentic data analysis. Code and data will be released at https://github.com/zjunlp/DataMind.

Related papers

• SPIEval: Evaluating Large Language Models as Mobile Assistants over Scattered Personal Information [61.41881137371855]
  We introduce SPIEval, a benchmark for large language models (LLMs)<n> SPIEval comprises 250 tasks spanning 4,335 personal records distributed across 10 apps and supports multi-turn interaction through 21 tools.<n>The best-performing model, GPT-5.5 (xhigh), achieves only 57.3% accuracy, while the weakest achieves just 16.4%.
  arXiv  Detail & Related papers  (2026-08-11T09:14:53Z)
• SANA: What Matters for QA Agents over Massive Data Lakes? [21.461240255601684]
  We present SANA, a framework that transforms EQA tasks into runtime profiles containing gold source sequence, sanitized subquestions, and execution records.<n>We evaluate two recent EQA benchmarks, LakeQA and KramaBench, and evaluate lightweight and mid-sized agents under fixed prompts, budgets, data lakes, and runtimes.
  arXiv  Detail & Related papers  (2026-06-11T20:51:44Z)
• DataClaw: A Process-Oriented Agent Benchmark for Exploratory Real-World Data Analysis [76.98578575566184]
  DataClaw is a process-oriented benchmark for exploratory real-world data analysis.<n>It contains approximately 2.06 million real-world records across enterprise, industry and policy domains.<n>DataClaw measures how far an agent progresses and where its reasoning breaks down.
  arXiv  Detail & Related papers  (2026-05-04T11:57:09Z)
• The Long-Horizon Task Mirage? Diagnosing Where and Why Agentic Systems Break [69.91172974854422]
  HORIZON is a cross-domain diagnostic benchmark for analyzing long-horizon failure behaviors in large language model (LLM) agents.<n>We evaluate state-of-the-art (SOTA) agents from multiple model families to study horizon-dependent degradation patterns.<n>Our findings offer an initial methodological step toward systematic, cross-domain analysis of long-horizon agent failures.
  arXiv  Detail & Related papers  (2026-04-13T19:11:42Z)
• AgenticRAGTracer: A Hop-Aware Benchmark for Diagnosing Multi-Step Retrieval Reasoning in Agentic RAG [7.139631028105273]
  We introduce AgenticRAGTracer, a benchmark for agent-based multi-hop reasoning.<n>It is primarily constructed by large language models and designed to support step-by-step validation.<n>Our benchmark spans multiple domains, contains 1,305 data points, and has no overlap with existing mainstream benchmarks.
  arXiv  Detail & Related papers  (2026-02-22T10:55:21Z)
• daVinci-Agency: Unlocking Long-Horizon Agency Data-Efficiently [35.39097522391409]
  Large Language Models (LLMs) excel at short-term tasks, scaling them to long-horizon agentic synthesis remains challenging.<n>We propose daVinci-Agency, which systematically mines structured supervision from chain-of-PRs.<n>DaVinci-Agency's PR-grounded structure inherently preserves the causal dependencies and iterative refinements essential for teaching persistent goal-cycle behavior.
  arXiv  Detail & Related papers  (2026-02-02T13:23:39Z)
• LongDA: Benchmarking LLM Agents for Long-Document Data Analysis [55.32211515932351]
  LongDA targets real-world settings in which navigating long documentation and complex data is the primary bottleneck.<n>LongTA is a tool-augmented agent framework that enables document access, retrieval, and code execution.<n>Our experiments reveal substantial performance gaps even among state-of-the-art models.
  arXiv  Detail & Related papers  (2026-01-05T23:23:16Z)
• PublicAgent: Multi-Agent Design Principles From an LLM-Based Open Data Analysis Framework [5.863391019411233]
  Large language models show promise for individual tasks, but end-to-end analytical expose fundamental limitations.<n>We present PublicAgent, a multi-agent framework that addresses these limitations through decomposition into specialized agents for intent clarification, dataset discovery, analysis, and reporting.
  arXiv  Detail & Related papers  (2025-11-04T21:48:11Z)
• DABstep: Data Agent Benchmark for Multi-step Reasoning [2.6709582216950767]
  DABstep is a novel benchmark for evaluating AI agents on realistic multi-step data analysis tasks.<n>It comprises over 450 real-world challenges derived from a financial analytics platform.<n>Dabstep is released with a public leaderboard and toolkit to accelerate research in autonomous data analysis.
  arXiv  Detail & Related papers  (2025-06-30T10:49:21Z)
• DSBench: How Far Are Data Science Agents from Becoming Data Science Experts? [58.330879414174476]
  We introduce DSBench, a benchmark designed to evaluate data science agents with realistic tasks.<n>This benchmark includes 466 data analysis tasks and 74 data modeling tasks, sourced from Eloquence and Kaggle competitions.<n>Our evaluation of state-of-the-art LLMs, LVLMs, and agents shows that they struggle with most tasks, with the best agent solving only 34.12% of data analysis tasks and achieving a 34.74% Relative Performance Gap (RPG)
  arXiv  Detail & Related papers  (2024-09-12T02:08:00Z)
• DACO: Towards Application-Driven and Comprehensive Data Analysis via Code Generation [83.30006900263744]
  Data analysis is a crucial analytical process to generate in-depth studies and conclusive insights. We propose to automatically generate high-quality answer annotations leveraging the code-generation capabilities of LLMs. Our DACO-RL algorithm is evaluated by human annotators to produce more helpful answers than SFT model in 57.72% cases.
  arXiv  Detail & Related papers  (2024-03-04T22:47:58Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.