The University of Hong Kong (XLANG Lab)
We track 23 benchmarks that The University of Hong Kong (XLANG Lab) made or helped make. Its models are on its lab page.
5 in use 22 with public code 20 with a leaderboard here
Benchmarks
| Name | Kind | Released | Best score | Status |
|---|---|---|---|---|
| OSWorld-Verified 369 open-ended tasks in real desktop apps. Agent acts through screenshots, mouse and keyboard. Execution-based checkers score success. | Desktop apps | Jul 2025 | 86% | In use |
| OSWorld 2.0 Agents complete 108 long-horizon workflows across 31 self-hosted websites and desktop apps, averaging 318 tool calls. Scored by binary completion and checkpoint partial credit. | Desktop apps | Jun 2026 | 20.6% | In use |
| OSWorld Agents complete 369 real computer tasks in Ubuntu desktop apps through screenshots and mouse and keyboard actions. Execution-based scripts check the final state. | Desktop apps | Apr 2024 | 72.1% | Fading |
| Bird-SQL Models convert natural language questions into SQL over 95 large real databases across 37 domains. Scored by execution accuracy. | Data science and SQL | May 2023 |  63.6% | Fading |
| OSWorld 2.1 Agents complete 108 long-horizon workflows across 31 self-hosted websites and desktop apps, averaging 318 tool calls. Scored by binary completion and checkpoint partial credit. | Desktop apps | Sep 2026 | 81.8% | In use |
| OSWorld-G Models map 564 desktop instructions to the right screen element or point, covering text matching, element recognition, layout understanding and fine manipulation. | Finding things on screen | May 2025 |  71.3% | Fading |
| OSWorld 2.1 (offline subset) Agents complete the 82 OSWorld 2.1 tasks on the official offline list, with no internet access. Scored by checkpoint partial credit and strict pass rate. | Desktop apps | Sep 2026 | 89.3% | In use |
| UniClawBench Agents act as proactive assistants on real-world tasks with live tools across many turns. Tasks are grouped by the capability they test, so failures can be traced to one skill. | General assistants | Jul 2026 | 47.5% | In use |
| Math-VR Models answer 5,000 English and Chinese math questions whose solutions need visual reasoning, such as drawing auxiliary lines or plotting functions. Answers are checked against ground truth. | Visual math and diagrams | Oct 2025 | 66.8% | System cards only |
| Spider 2.0 Models and agents write SQL for real enterprise databases with large schemas, across BigQuery, Snowflake, SQLite and dbt projects. Answers are checked by executing them. | Data science and SQL | Nov 2024 |  94.2% | Dropped |
| OSWorld 2.0 (Aug 2026) Agents complete 108 long-horizon workflows across 31 self-hosted websites and desktop apps, averaging 318 tool calls. Scored by binary completion and checkpoint partial credit. | Desktop apps | Aug 2026 | 77.9% | System cards only |
| EvoCode-Bench Coding agents keep one project working across 26 tasks of 5 to 15 rounds each, while requirements change, pile up and conflict. Cumulative hidden tests check every round in a persistent container. | Building whole projects | May 2026 | 54% | Not in cards |
| EvoCode-Bench v2 Coding agents work through 26 multi-round tasks with 227 graded rounds on the official Harbor multi-step format. The score is the mean of binary per-round rewards. A case score gives partial credit. | Building whole projects | Jun 2026 | 59.1% | Not in cards |
| CineRecon Agents use tools to rebuild reference videos from film, real footage and games as editable dynamic 3D scenes. The set has 50 videos and 360 evaluation clips. | Video understanding | Oct 2026 | 76.6 | Not in cards |
| VL-RewardBench Vision-language reward models pick the better of two responses on general queries, hallucination detection and reasoning tasks. | General visual questions | Nov 2024 |  53.9% | Not in cards |
| Spider2-V Multimodal agents do data science and engineering workflows in real enterprise apps such as BigQuery, dbt and Airbyte, on a desktop computer. Execution-based checks score each task. | Desktop apps | Jul 2024 | 14% | Not in cards |
| DS-1000 Models write Python for 1,000 data science problems from StackOverflow across seven libraries such as NumPy and Pandas. Test cases and API constraints check each answer. | Data science and SQL | Nov 2022 | 53.9% | Not in cards |
| DriveBench Behavior Models answer driving behavior questions. The score is the published GPT evaluation score for this task family. | Embodied reasoning | Jan 2025 |  54.6 | Not in cards |
| DriveBench Perception Models answer driving perception questions. The score is the published GPT evaluation score for this task family. | Embodied reasoning | Jan 2025 | 35.2 | Not in cards |
| DriveBench Planning Models answer driving planning questions. The score is the published GPT evaluation score for this task family. | Embodied reasoning | Jan 2025 | 57.0 | Not in cards |
| DriveBench Prediction Models answer driving prediction questions. The score is the published GPT evaluation score for this task family. | Embodied reasoning | Jan 2025 | 48.9 | Not in cards |
| RMBench Policies run 9 simulated bimanual manipulation tasks on RoboTwin 2.0 that need memory of earlier observations, at several levels of memory complexity. Success rate. | Simulated robots | Mar 2026 | 10.4% | Not in cards |
| DriveBench Vision-language models answer 20,498 questions about 19,200 driving frames, covering perception, prediction, planning and behavior. Inputs come in 17 settings: clean images, corrupted images and text only. The test checks whether answers rest on the image or on general knowledge. | Embodied reasoning | Jan 2025 | Not in cards |