CritPt
Harbor: pendingCritPt — AI evaluation benchmark listed by opencompass.
Verified71%Knowledge & Reasoning
- Samples
- 70
- Released
- 2025
- Scores
- 138
GeopoliticalQA
GeopoliticalQA: Benchmarking Retrieval and Generation over Intelligence Reports
A benchmark for grounded QA over 415 geopolitical intelligence reports, comprising 5,525 document-grounded question-answer pairs spanning factual, inferential, comparative, and risk-oriented questions.
How to run
ltb · geopoliticalqa
$ npm i -g lettokenburn-cli$ ltb bench info geopoliticalqa
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
6 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| Llama-3.3-70B-Instruct | 2.6 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| Mistral-Small-3.2 | 2.5 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| Qwen3-32B | 2.5 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| Qwen3 8B | 2.5 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| Llama-3.1-8B-Instruct | 2.3 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| Mistral NeMo | 2.3 | N/A | Independent | medium | Oct 8, 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 dataset contains 5,525 QA pairs derived from 415 open-source geopolitical reports using an automated, document-aware generation pipeline. It covers diverse institutional sources and question types including factual and risk-oriented queries.
Evaluation involves 15 retrieval systems and six LLMs, assessed via expert annotation and an LLM-as-judge protocol. Findings indicate chunk-level retrieval quality is not driven by embedding scale, while answer generation remains challenging for inferential tasks.