university collaborators
We track 9 benchmarks that university collaborators made or helped make.
0 in use 9 with public code 8 with a leaderboard here
Benchmarks
| Name | Kind | Released | Best score | Status |
|---|---|---|---|---|
| Long-Horizon-Terminal-Bench (LHTB) Terminal-Bench-style container tasks needing hundreds of steps. Subtasks graded deterministically for dense partial-credit reward. Solved at reward >=0.95. | Long terminal sessions | Jul 2026 | | Not in cards |
| Argus Inspection Models inspect images for small details that change the sensible response to a question. A judge scores whether the response identifies the relevant detail and chooses an appropriate action. The score combines those judgments and assigns zero when a separate description fails to recognize the detail. | General visual questions | Jun 2025 | | Not in cards |
| SpaceDG-Bench Models answer spatial questions under clean images and nine forms of visual degradation. Choice and binary answers use accuracy, while numerical answers use relative-error thresholds. The test has 1,102 verified questions from scenes excluded from the training set. | Spatial reasoning and counting | May 2026 | | Not in cards |
| Long-Horizon-Terminal-Bench 2.0 Agents work for up to 3 hours per task on 77 container tasks, including 30 tasks that rebuild a scientific tool from its outputs. Hidden verifiers give partial-credit rewards. | Long terminal sessions | Oct 2026 | 69.2% | Not in cards |
| GUI Knowledge Bench Models answer 3,483 questions about interface states, actions, and procedures. They inspect screenshots and task sequences from 292 applications across six platforms. | Finding things on screen | Oct 2025 | 78.9% | Not in cards |
| VDR-Bench Models answer 2,000 visual questions that require finding entities in images and researching related facts. A Qwen3-VL judge checks final answers against references. Entity recall separately measures whether searches retrieve the relevant entities. | Deep research | Feb 2026 | 27.2% | Not in cards |
| DEEPSYNTH Agents search multiple websites and combine evidence to answer 120 research tasks across seven domains. They return structured JSON answers scored with precision, recall, F1, and exact match. A separate model judge assesses answers beyond exact matching. | Deep research | Feb 2026 | 8.97% | Not in cards |
| OmniEarth-Bench Models interpret satellite images and observational data across Earth systems, answering questions about environmental processes and their interactions. Multiple-choice accuracy and model judgments score answers. Separate measures check localization, captions, and annotated reasoning steps. | Research science | May 2025 | | Not in cards |
| OpenHuEval Models answer Hungarian questions, interpret proverbs, and fill gaps in texts across five tasks. Factual questions and open responses use model judges, while matching and blank completion also use rules and text similarity. | Reasoning in many languages | Mar 2025 | Not in cards |