AgentHorizon
Harbor: pendingA benchmark of 1,373 computer-use tasks with paired instructions to evaluate agentic judges on long-horizon trajectory success verification across three OSs.
- Samples
- 1,373
- Released
- 2026
- Scores
- 10
KBS
Korean Benchmark Suite
A suite of 9,396 Korean items across six capabilities (science, math, sentiment, reasoning, context, instruction) evaluating foundation models.
How to run
ltb · kbs
$ npm i -g lettokenburn-cli$ ltb bench info kbs
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
4 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| GPT-5.5 | 96.3 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| Claude Opus 4.7 | 96.1 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| GPT-5-mini | 94.5 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
| Claude Haiku 4.5 | 93.5 | N/A | Independent | medium | Oct 15, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsLoading community runs…
About
KBS comprises Ko-ARC, Ko-GSM8K, Ko-EQ-Bench, Ko-WinoGrande, Ko-LAMBADA, and Ko-IFEval. It uses review, localization, and direct construction to avoid translation-only artifacts.
It evaluates Korean, multilingual, and closed models using protocols accounting for prompt format and scoring. Findings highlight dependence on task construction and suggest reporting both scores and settings.