AWARE-US (Local Weighting) - S1 Unique Repair: leaderboard
Setting S1 of AWARE-US (Kurmaz, 2026): 41 persona-grounded car-search dialogues with four preference constraints of which exactly one can be dropped to make the empty query feasible, where an LLM agent elicits the constraints, finds the query infeasible and must relax the constraint the persona values least. The agent uses the local-weighting method (each clarification answer yields a constraint and an importance weight, and the lowest-weight constraint is relaxed first). Car match is the share of dialogues in which both the relaxed constraint and the recommended car agree with the oracle; higher is better.
Metric: Car match (%; share of the 41 persona-grounded car-search dialogues, four preference constraints of which exactly one can be dropped to make the query feasible, in which the agent relaxes the oracle's least important constraint and recommends the oracle's car; local weighting: each clarification answer yields a constraint and an importance weight, and the lowest-weight constraint is relaxed first). Source: arxiv.org. Saturation forecast: Estimated already saturated. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 2 9B (IT) | 46.3 | #2275 |
| 2 | Qwen 3 8B | 43.9 | #1838 |
| 3 | Llama 3 8B Instruct | 14.6 | #4652 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=aware-us-local-weighting-s1-unique-repair · How It Works · Data refreshed daily, snapshot 2026-11-12.