KIDBench - Cultural Alignment (Pakistan): leaderboard

KIDBench Cultural Alignment Pakistan rates whether LLM answers to children aged 7 to 11 adapt guidance, help-seeking pathways and social expectations to Pakistan, when the system instruction names the child age and country (KIDBench, 2026). A DeepSeek-V4-Pro judge with country-specific rules scores each answer from 1 to 5, averaged over ten categories; higher is better.

Metric: Cultural Alignment score (1-5): whether the response adapts guidance, help-seeking pathways and social expectations to Pakistan, rated by a DeepSeek-V4-Pro judge with country-specific cultural rules (temperature 0), on the 500 implicit-cue KIDBench prompts in English with a system instruction naming a child aged 7 to 11 in Pakistan, mean over the 10 categories, single-turn; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3.6 27B4.22#358
2Gemma 3 12B4.04#2091
3Gemini 3.1 Flash Lite3.94#377
4GPT-5 Mini3.88#373
5Gemma 4 31B (IT)3.81#354
6DeepSeek V4 Flash3.8#273
7Claude Haiku 4.53.57#587
8Gemma 3 4B3.27#5347
9Qwen 3.5 4B3.18#1270
10Llama 3.3 70B Instruct3.01#1394
11Qwen 3 8B2.98#1890
12Llama 3.1 8B Instruct2.64#4655
13KIDBench Llama-3.2-3B (checkpoint unspecified)2.56

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=kidbench-cultural-alignment-pakistan · How It Works · Data refreshed daily, snapshot 2026-11-22.