AIME 2025
Harbor adapterAIME 2025: 30道AIME级数学题
Verified86%Math
- Samples
- 30
- Released
- 2025
- Scores
- 12
Colombian Legal Reliability Benchmark
A Reliability Benchmark for the Colombian Legal System
Expert-validated benchmark of 1042 items across ten legal areas and three formats (MCQ, semi-open, IRAC) evaluating LLM reliability on Colombian law.
How to run
ltb · colombian-legal-reliability-benchmark
$ npm i -g lettokenburn-cli$ ltb bench info colombian-legal-reliability-benchmark
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
13 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| Gemini 3.1 Pro | 90.5 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Claude Opus 4.5 (2025-11-01) | 86.9 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| GPT-5.4 | 85.9 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| GLM-5.2 | 85.6 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Claude Sonnet 4.5 (2025-09-29) | 80.3 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| GLM-4.6 | 79.0 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Kimi K2.5 | 75.4 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| DeepSeek V3.2 | 72.8 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| WizardLM-2-8x22B | 72.1 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Qwen3 235B | 70.8 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Nemotron 49B | 69.5 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Qwen2.5-72B-Instruct | 67.2 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Llama-3.3-70B-Instruct | 66.2 | N/A | Independent | medium | Oct 2, 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
The benchmark comprises 1042 expert-validated questions spanning ten areas of Colombian law, built via a human-in-the-loop pipeline. It includes closed multiple-choice, semi-open, and open-ended (IRAC) formats.
Scoring uses format-appropriate metrics, including accuracy for MCQs and factual correctness for free text. Results show high MCQ accuracy but low free-text correctness (<0.45), revealing a dissociation between relevancy and correctness.