Chinese contentWorld 模型排序策略不必完全准确：相对误差界与审计极限
这篇 arXiv 论文提出 RAVEL 方法，用 world 模型低成本给 RL 策略排序，淘汰阶段再用少量真实审计兜底，还量化了模型排序可靠到什么程度、什么时候会失灵。
#RAVEL
2 stories · 1 in 7 days · 1 in 30 days
Stories, products and papers tagged "RAVEL", newest first.
Nov 3
Sep 21
RAVEL Learns Retrieval-Aware Questioning for Interactive Person Re-identification
The framework replaces imitation of offline QA orderings with online rank feedback, giving practitioners a concrete recipe for training question-asking retrieval agents on benchmarks like Interactive-PEDES.