Tencent's LifeSearch benchmark: best deep research agent scores just 63.4% on real-life tasks
Tencent-Hunyuan · hf · 2026-10-14
Tencent Hunyuan released LifeSearch, a benchmark for evaluating deep research agents on real-world life and productivity tasks, filling a gap left by academic/synthetic benchmarks.
- 203 Chinese and English tasks from real-user query seeds across 7 domains; 4,500+ hours of annotation produced 4,000+ expert-written rubrics covering report quality and concrete values, units, dates, and URLs
- Under a unified harness, 12 advanced DR agents were tested — the best, GPT-5.6 Sol, scored only 63.4%
- LLM judge configuration calibrated against human labels; scores align strongly with human rankings (Spearman ρ=0.89, Kendall τ=0.81)
- Detailed analysis of performance patterns and failure modes
More from Research
- Swaayatt Demos On-the-Fly Camera-LiDAR Calibration Without Markers in the Wild — sanjeevs_iitr · 2026-10-14
- LongR: RL with dense utility rewards boosts long-context reasoning by 9% on LongBench v2 — PKU · 2026-10-14
- On-policy distillation study: 4 prompts match 3,840, but prompt utility depends on the teacher-student pair — Tencent-Hunyuan · 2026-10-14
- RISC-V crypto accelerator SoC gets full place-and-route done with ~$2,000 of LLM tokens — IgorCarron · 2026-10-14
- Tufts and Santa Fe Institute paper models LLM adoption as a 'cognitive virus' — alex_verem · 2026-10-14
- PyVBBC 1.5 Released After 3 Years, Bringing Major Upgrades to Bayesian Model Inference — twiecki · 2026-10-14