HPR-Bench - Tag Inference - Household Context: leaderboard
Household attributes (the number, ages and genders of a user's children) inferred from a long multi-source behavioural history in HPR-Bench (UserGPT Technical Report, Xuan et al., 2026), asked as single- or multiple-choice questions. Avg@10 accuracy, averaged over ten sampled answers, in percent; higher is better.
Metric: Avg@10 Tag Accuracy (%; mean of 10 samples). Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.6 Plus (Thinking) | 59.65 | #217 (Qwen 3.6 Plus) |
| 2 | Qwen 3.5 397B A17B (Thinking) | 54.66 | #242 (Qwen 3.5 397B A17B) |
| 3 | Qwen 3.5 Plus (Thinking) | 54 | #227 (Qwen 3.5 Plus) |
| 4 | Kimi K2 (Thinking) | 53.8 | #435 (Kimi K2) |
| 5 | Kimi K2.5 (Thinking) | 53.8 | #252 (Kimi K2.5) |
| 6 | GLM-5 (Thinking) | 53.45 | #256 (GLM-5) |
| 7 | GLM-4.7 (Thinking) | 52 | #365 (GLM-4.7) |
| 8 | DeepSeek V3.2 (Thinking) | 49.81 | #403 (DeepSeek V3.2) |
| 9 | DeepSeek R1 0528 | 47.52 | #425 |
| 10 | Qwen 3 235B A22B 2507 Instruct | 41.4 | #619 |
| 11 | Qwen 3 235B A22B 2507 (Thinking) | 41.35 | #460 |
| 12 | Qwen 3 8B (Thinking) | 38.31 | #1838 (Qwen 3 8B) |
| 13 | Qwen 3 8B (Non-reasoning) | 32.81 | #1838 (Qwen 3 8B) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=hpr-bench-tag-inference-household-context · How It Works · Data refreshed daily, snapshot 2026-11-12.