ALPBench - Baijiu (Profile F1): leaderboard
Metric: Profile-level F1 (%): the harmonic mean of the shares of predicted and of ground-truth attribute combinations that match exactly (every attribute right), for the next purchased baijiu item's attribute combination (price range, brand, alcohol content and packaging), predicted zero-shot from 100 users' one-year Kuaishou e-commerce interaction histories (clicks, add-to-cart and purchases with titles, selling points and price tiers, up to about 340K tokens); ground truth from verified purchases after a shopping festival cut-off; higher is better. Source: arxiv.org. Saturation forecast: Around 2028. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seed 1.8 | 20.08 | #136 |
| 2 | Gemini 2.5 Pro | 19.2 | #145 |
| 3 | DeepSeek V3.2 | 18.87 | #198 |
| 4 | Claude Sonnet 4.5 | 18.45 | #138 |
| 5 | Qwen 3 Max (Preview) | 17.8 | #224 |
| 6 | GLM-4.5 | 17.63 | #265 |
| 7 | Gemini 2.5 Flash | 17.53 | #237 |
| 8 | GLM-4.6 | 17.07 | #246 |
| 9 | MiniMax-M2 | 17.02 | #307 |
| 10 | Qwen 3 235B A22B 2507 Instruct | 16.9 | #291 |
| 11 | MiniMax-M2.1 | 16.57 | #283 |
| 12 | DeepSeek R1 | 15.83 | #245 |
| 13 | GPT-5.2 Instant | 15.37 | #205 |
| 14 | Qwen 3 30B A3B 2507 (Thinking) | 13.67 | #366 (Qwen 3 30B A3B 2507) |
| 15 | Kimi K2 | 6.67 | #236 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=alpbench-baijiu-profile-f1 · How It Works · Data refreshed daily, snapshot 2026-10-11.