# PANDO：通过在线技能蒸馏实现高效多模态AI智能体

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-05-26 08:00
- AIHOT 分数：54
- AIHOT 链接：https://aihot.virxact.com/items/cmprhxl4z016lslljmqbacm7e
- 原文链接：https://arxiv.org/abs/2605.24785

## AI 摘要

多模态网页智能体 PANDO 旨在解决推理计算成本随经验增长的问题。它通过分析 VisualWebArena 轨迹，识别出重复动作循环等低效来源，并提出了单轮在线技能蒸馏框架。PANDO 维护结构化技能库，结合反思、分层路由、视觉压缩等技术。在 910 个任务上，其成功率达 58.3%，优于 SGV (54.0%) 和 WALT (45.2%)，且 token 消耗分别减少 58% 和 61%，无需预评估预算。消融研究也验证了其高效性。

## 正文

Recent advances in multimodal web agents often rely on increased inference-time computation, including rollout search, verifier passes, offline skill discovery, and specialist model stacks. This raises a central question: can a web agent become more efficient as it accumulates experience, rather than more expensive? We first analyze trajectories from VisualWebArena and identify three recurring sources of inefficiency: repeat-action loops, hidden discovery costs, and low prompt-cache reuse. We then introduce PANDO, a single-rollout online skill-distillation framework that maintains a structured Skill Library and combines progress reflection, confidence-based skill demotion, hierarchical routing, visual compression, and cache-aware prompting. On the full set of 910 VisualWebArena tasks, PANDO achieves a 58.3% success rate, outperforming SGV (54.0%) and our WALT reproduction (45.2%), while using 58% fewer tokens than SGV and 61% fewer tokens than WALT, without any pre-evaluation discovery budget. A 300-task ablation further shows that rules and routines provide most of the success gains, while routing, compression, and cache-aware prompting convert the larger skill library into lower marginal token cost. Finally, we introduce three trajectory-level efficiency metrics -- Action Repetition Rate, Step Overhead Ratio, and Prompt Cache Utilization -- to make efficiency visible beyond terminal success.