# LH-Bench: Skill-Grounded Evaluation of Long-Horizon Agents on Subjective Enterprise Tasks

arXiv: https://arxiv.org/abs/2603.22744 · HTML: https://mlsanity.com/paper/2603.22744/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use.
- Builds on 19 of its 31 references that are in the map.
- First posted on arXiv 24 Mar 2026.

## Placed in

- LLM agents and tool use / [General LLM agent frameworks and benchmarks](https://mlsanity.com/agents/general-llm-agent-frameworks-and-benchmarks/)

## Builds on (most cited first, 19 of 19)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [WebArena: A Realistic Web Environment for Building Autonomous Agents](https://mlsanity.com/paper/2307.13854/) (2023)
- [ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs](https://mlsanity.com/paper/2307.16789/) (2023)
- [A Survey on LLM-as-a-Judge](https://mlsanity.com/paper/2411.15594/) (2024)
- [τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains](https://mlsanity.com/paper/2406.12045/) (2024)
- [OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments](https://mlsanity.com/paper/2404.07972/) (2024)
- [GAIA: a benchmark for General AI Assistants](https://mlsanity.com/paper/2311.12983/) (2023)
- [τ2-Bench: Evaluating Conversational Agents in a Dual-Control Environment](https://mlsanity.com/paper/2506.07982/) (2025)
- [WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?](https://mlsanity.com/paper/2403.07718/) (2024)
- [MINT: Evaluating LLMs in Multi-turn Interaction with Tools and Language Feedback](https://mlsanity.com/paper/2309.10691/) (2023)
- [SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?](https://mlsanity.com/paper/2509.16941/) (2025)
- [Design2Code: Benchmarking Multimodal Code Generation for Automated Front-End Engineering](https://mlsanity.com/paper/2403.03163/) (2024)
- [ResearchRubrics: A Benchmark of Prompts and Rubrics For Evaluating Deep Research Agents](https://mlsanity.com/paper/2511.07685/) (2025)
- [DeepResearch Bench II: Diagnosing Deep Research Agents via Rubrics from Expert Reports](https://mlsanity.com/paper/2601.08536/) (2026)
- [Multi-Agent-as-Judge: Aligning LLM-Agent-Based Automated Evaluation with Multi-Dimensional Human Evaluation](https://mlsanity.com/paper/2507.21028/) (2025)
- [FullFront: Benchmarking MLLMs Across the Full Front-End Engineering Workflow](https://mlsanity.com/paper/2505.17399/) (2025)
- [UltraHorizon: Benchmarking Agent Capabilities in Ultra Long-Horizon Scenarios](https://mlsanity.com/paper/2509.21766/) (2025)
- [FrontendBench: A Benchmark for Evaluating LLMs on Front-End Development via Automatic Evaluation](https://mlsanity.com/paper/2506.13832/) (2025)
- [FronTalk: Benchmarking Front-End Development as Conversational Code Generation with Multi-Modal Feedback](https://mlsanity.com/paper/2601.04203/) (2025)

## Built on by (most cited first, 3 of 3)

- [Terminal Agents: A Survey of AI Agents in Command-Line Environments](https://mlsanity.com/paper/2608.20485/) (2026)
- [FigmaTrace: Capturing Creative Nuances in Human Figma Design Workflows](https://mlsanity.com/paper/2608.21460/) (2026)
- [World Potential Model: Pretrained World Knowledge as Progress Potentials](https://mlsanity.com/paper/2610.09560/) (2026)

## Abstract (arXiv)

Large language models excel on objectively verifiable tasks such as math and programming, where evaluation reduces to unit tests or a single correct answer. In contrast, real-world enterprise work is often subjective and context-dependent: success hinges on organizational goals, user intent, and the quality of intermediate artifacts produced across long, multi-tool workflows. We introduce LH-Bench, a three-pillar evaluation design that moves beyond binary correctness to score autonomous, long-horizon execution on subjective enterprise tasks. The pillars are: (i) expert-grounded rubrics that give LLM judges the domain context needed to score subjective work, (ii) curated ground-truth artifacts that enable stepwise reward signals (e.g., chapter-level annotation for content tasks), and (iii) pairwise human preference evaluation for convergent validation. We show that domain-authored rubrics provide substantially more reliable evaluation signals than LLM-authored rubrics (kappa = 0.60 vs. 0.46), and that human preference judgments confirm the same top-tier separation (p < 0.05), evidence that expert-grounded evaluation can scale without sacrificing reliability. We release public datasets and report results on two environments: Figma-to-code (33 real .fig tasks against the Figma API via MCP) and Programmatic content (41 courses comprising 183 individually-evaluated chapters on a course platform serving 30+ daily users).