TravelEval - Headcount Deviation: leaderboard
TravelEval (KDD 2026) asks an LLM for a complete multi-day travel plan and simulates it in a sandbox of 10 Chinese cities with real transport, hotel and attraction data. Scores are from Direct prompting over 1,150 queries. The board is the summed gap between booked beds, seats and tickets and the party size, divided by the party size; lower is better.
Metric: Personnel Deviation (PD): summed absolute deviation of beds, vehicle seats and tickets from the party size, over the party size (ratio, 0 is exact); Direct prompting (one pass, no tools), plans simulated in the TravelEval sandbox of 10 Chinese cities (real rail, flight, hotel and attraction data, queuing-time model, road-network distances) over 1,150 queries; lower is better. Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 0.3 | #250 |
| 2 | Qwen 3 8B | 0.6 | #1830 |
| 3 | DeepSeek V3.1 (Non-reasoning) | 0.62 | #497 (DeepSeek V3.1) |
| 4 | Gemini 2.0 Flash | 0.64 | #704 |
| 5 | GPT-4o Mini (2024-07-18) | 0.65 | #1510 |
| 6 | GPT-5 Chat | 0.71 | #444 |
| 7 | GPT-4o (2024-11-20) | 0.74 | #866 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=traveleval-headcount-deviation · How It Works · Data refreshed daily, snapshot 2026-11-13.