LetTokenBurnUXBench Pro: Benchmarking Personalized User Experience in Multi-Turn Dialogue Interactions
A benchmark of 1,000 multi-turn dialogue instances with FACTORS user profiles across 12 scenarios and 82 domains, evaluating personalized UX via reward models and simulators.
$ npm i -g lettokenburn-cli$ ltb bench info uxbench-pro
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| Qwen3.7 Max | 0.2 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| DeepSeek V4 Pro | 0.0 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| Hunyuan-3 | 0.0 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| GLM-5.2 | 0.0 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| Claude Opus 4.8 | -0.0 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| GPT-5.5 | -0.1 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
| Gemini 3.1 Pro | -0.1 | N/A | Independent | medium | Oct 8, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
UXBench Pro comprises 1,000 test instances from real user interactions, each paired with a seven-facet FACTORS user profile. It covers 12 task scenarios and 82 domains to capture user heterogeneity.
Evaluation uses a dual-perspective paradigm: a User Reward Model for third-person judgment and Sim4Eval for first-person multi-turn assessment. Meta-benchmarks URMBench and USimBench validate evaluator reliability against human preferences.