跳到正文
arXiv cs.AI· Eleftherios Triantafyllidis, Filippos Christianos, Zhibin Li, Bernd Bickel·· 2 小时前AI 评分39

WATERFALL:面向 POMDP 环境自动奖励发现的进化式训练工作流

WATERFALL: Workflow for Adaptive Training with Evolutionary Reward Formulation and Automated Learning Loops

AI 导读

WATERFALL 是一个在 POMDP 部分可观测环境下进行自动奖励发现的迭代式、基于种群的工作流。它通过 persona 条件化生成程序化奖励候选,用视觉语言模型评判的瑞士制进化锦标赛从原始视觉 rollout 中筛选精英,并借助纵向评估历史迭代变异或淘汰候选。

正文 · 原文

View PDF HTML (experimental)

Abstract:Reward design remains a central bottleneck in reinforcement learning (RL), particularly in sparse-reward, long-horizon and partially observable settings requiring dependency chaining, memory, and deceptive affordance disambiguation. While recent foundation-model approaches reduce manual reward engineering, most assume goal-conditioning, task descriptions, expert demonstrations, or ground-truth metrics. These assumptions are problematic in partially observable environments, where task-relevant objectives, affordances, and dynamics must be discovered through interaction rather than disclosed a priori. We formalise this problem setting by introducing WATERFALL, an iterative population-based workflow for automated reward discovery under the strict observability constraints inherent in POMDP environments. WATERFALL achieves this by (i) synthesising diverse programmatic reward candidates via persona-conditioned generation, (ii) evaluating candidate behaviour from raw visual rollouts utilising a Swiss-system evolutionary tournament judged by vision-language models to identify elites, and (iii) leveraging longitudinal assessment histories to iteratively mutate, escalate, or prune candidates. We evaluate WATERFALL across fully observable and reveal-gated partially observable environments ranging from discrete to continuous, comparing against sparse-reward RL, intrinsic-motivation and foundation-model reward-synthesis baselines. Our results indicate that WATERFALL discovers and refines reward programs without access to privileged information, outperforming baselines that retain their native information advantages, most clearly in the POMDP regime. Extensive ablations substantiate that visual evidence, tournament-based selection, and iterative refinement are jointly important. Enforcing this contract is what separates autonomous reward discovery from benchmark-privileged reward shaping.
Comments: Accepted at NeurIPS 2026 (poster). 78 pages, 13 figures, 19 tables. Code: this https URL OpenReview: this https URL
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as: arXiv:2611.22854 [cs.LG]
  (or arXiv:2611.22854v1 [cs.LG] for this version)
  https://doi.org/10.48550/arXiv.2611.22854

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Eleftherios Triantafyllidis [view email]
[v1] Tue, 17 Nov 2026 14:35:27 UTC (19,706 KB)

来源:arXiv cs.AI · arxiv.org