CritPt
Harbor: pendingCritPt — AI evaluation benchmark listed by opencompass.
Verified71%Knowledge & Reasoning
- Samples
- 70
- Released
- 2025
- Scores
- 141
ChronoCulture
ChronoCulture: Evaluating Cultural Awareness Across Time in Large Language Models
A benchmark of 10,360 temporal pairs from longitudinal surveys across 72 countries, evaluating if LLMs capture cultural attitude changes over time using full response distributions.
How to run
ltb · chronoculture
$ npm i -g lettokenburn-cli$ ltb bench info chronoculture
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 | |||
|---|---|---|---|---|---|---|
| GPT-5.4 | 80.3 | N/A | Independent | medium | Oct 14, 2026 | arxiv.org |
| Qwen3.8 Flash | 68.3 | N/A | Independent | medium | Oct 14, 2026 | arxiv.org |
| GPT-4o | 66.9 | N/A | Independent | medium | Oct 14, 2026 | arxiv.org |
| Llama-3.3-70B-Instruct | 65.1 | N/A | Independent | medium | Oct 14, 2026 | arxiv.org |
| DeepSeek-V3 | 60.5 | N/A | Independent | medium | Oct 14, 2026 | arxiv.org |
| Qwen-3.5-35B | 54.3 | N/A | Independent | medium | Oct 14, 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
ChronoCulture comprises 10,360 temporal pairs derived from World Values Survey and ISSP data across 72 countries (1985-2022). Each pair presents the same survey question for a specific country in two different years, with weighted population response distributions as ground truth.
Models predict response distributions for each year. Evaluation assesses alignment with observed distributions and the ability to capture cultural change rather than static snapshots. Findings indicate LLMs often fail to reflect temporal shifts in attitudes.