LetTokenBurnLifeSearch: Benchmarking Deep Research Agents for Real-World Life and Productivity
A bilingual benchmark of 203 real-world tasks with 4,000+ expert rubrics evaluating deep research agents on life and productivity domains.
$ npm i -g lettokenburn-cli$ ltb bench info lifesearch
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 | |||
|---|---|---|---|---|---|---|
| GPT-5.6 Sol | 63.4 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Claude Opus 5 | 56.1 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| HY4 Preview | 49.7 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| GLM-5.3 | 49.2 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Qwen3.8 Max | 47.5 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| DeepSeek V4 Pro | 46.9 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Kimi K3 | 45.8 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Doubao-Seed-2.1-Pro | 43.1 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Claude Opus 4.8 | 41.9 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Hy3 | 38.9 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| GLM-5.2 | 37.1 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
| Gemini 3.6 Flash | 34.7 | N/A | Independent | medium | Oct 13, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
LifeSearch comprises 203 Chinese and English tasks across 7 life and productivity domains, constructed from crowd-sourced seeds and refined by experts. It features over 4,000 fine-grained, numerically grounded rubrics assessing report coverage and accuracy.
Scoring uses an LLM-as-a-judge (Gemini 3.1 Pro) aligned with human rankings. Experiments on 12 agents show a top score of 63.4%, indicating significant headroom for improvement in real-world deep research capabilities.