CreativePreferences
Harbor: pendingA dataset of 2.8M texts with 317M human preference judgments across 7 creative domains, featuring 42 benchmark tasks for preference prediction.
- Samples
- 2,800,000
- Released
- 2026
- Scores
- 7
ViLegalExpert
ViLegalExpert: A Large-Scale Benchmark for Vietnamese Legal Retrieval and Question Answering from Real-World Consultations
A large-scale benchmark containing over 172K authentic legal questions across 34 domains, paired with professional answers and expert-verified evidence to evaluate retrieval and QA.
How to run
ltb · vilegalexpert
$ npm i -g lettokenburn-cli$ ltb bench info vilegalexpert
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
5 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| LLama-3 8B | 73.0 | N/A | Independent | medium | Sep 30, 2026 | arxiv.org |
| Llama-3.1-8B-Instruct | 72.1 | N/A | Independent | medium | Sep 30, 2026 | arxiv.org |
| Qwen2.5-7B-Instruct | 71.3 | N/A | Independent | medium | Sep 30, 2026 | arxiv.org |
| Qwen2.5 14B Instruct | 70.1 | N/A | Independent | medium | Sep 30, 2026 | arxiv.org |
| Mistral-7B-Instruct-v0.3 | 64.0 | N/A | Independent | medium | Sep 30, 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
Content & Construction: ViLegalExpert is constructed from real-world online consultations between citizens and lawyers in Vietnam. It includes a corpus of authoritative government documents (Constitution, laws, decrees) linked via expert annotation to the consultation questions.
Scoring & Findings: The benchmark supports legal information retrieval, extractive QA, and abstractive QA. Experiments reveal substantial challenges in mapping natural language queries to specific provisions; hybrid retrieval performs best for evidence finding.