视频选题榜 2026-07-19
- https://arxiv.org/abs/2607.12227 · Rethinking the Evaluation of Harness Evolution for Agents · 自动 harness 演化 ≠ scaling:留出任务上几乎不优于 baseline · round=R1
- https://arxiv.org/abs/2603.21972 · Demystifying RL for Long-Horizon Tool-Using Agents: A Comprehensive Recipe · 5 维设计空间 + scale-dependent reward + 1K 样本甜点 · round=R2
- https://arxiv.org/abs/2607.14387 · Chat2Scenic · NL→DSL 编译反馈回灌 + 三源检索 · round=R3