The reviewed record of science sign in
Pith

arxiv: 2310.03720 · v4 · pith:2IE7GBSS · submitted 2023-10-05 · cs.LG

SteP: Stacked LLM Policies for Web Actions

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:2IE7GBSSrecord.jsonopen to challenge →

classification cs.LG
keywords policiesstepcontrollargetasksactionsbehaviorsdata
0
0 comments X
read the original abstract

Performing tasks on the web presents fundamental challenges to large language models (LLMs), including combinatorially large open-world tasks and variations across web interfaces. Simply specifying a large prompt to handle all possible behaviors and states is extremely complex, and results in behavior leaks between unrelated behaviors. Decomposition to distinct policies can address this challenge, but requires carefully handing off control between policies. We propose Stacked LLM Policies for Web Actions (SteP), an approach to dynamically compose policies to solve a diverse set of web tasks. SteP defines a Markov Decision Process where the state is a stack of policies representing the control state, i.e., the chain of policy calls. Unlike traditional methods that are restricted to static hierarchies, SteP enables dynamic control that adapts to the complexity of the task. We evaluate SteP against multiple baselines and web environments including WebArena, MiniWoB++, and a CRM. On WebArena, SteP improves (14.9\% to 33.5\%) over SOTA that use GPT-4 policies, while on MiniWob++, SteP is competitive with prior works while using significantly less data. Our code and data are available at https://asappresearch.github.io/webagents-step.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 5 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. DRIVE: Modeling Skills at the Reasoning and Interaction Levels for Web Agents under Continual Learning

    cs.AI 2026-04 unverdicted novelty 6.0

    DRIVE disentangles reasoning and interaction skills for web agents via dual-level modeling and scene-aware coordination, reaching 52.8% success on WebArena tasks.

  2. Agent Workflow Memory

    cs.CL 2024-09 unverdicted novelty 6.0

    AWM induces reusable workflows from agent experiences and provides them selectively to improve success rates by 24.6% on Mind2Web and 51.1% on WebArena while reducing steps taken.

  3. Plan-and-Act: Improving Planning of Agents for Long-Horizon Tasks

    cs.CL 2025-03 unverdicted novelty 5.0

    Plan-and-Act trains a dedicated Planner on synthetic plan-annotated trajectories to generate high-level plans that an Executor follows, reaching 57.58% success on WebArena-Lite and 81.36% on WebVoyager.

  4. A Comprehensive Survey of Agents for Computer Use: Foundations, Challenges, and Future Directions

    cs.AI 2025-01 unverdicted novelty 5.0

    A survey of 87 agents for computer use and 33 datasets that introduces a three-dimensional taxonomy across domain, interaction, and agent perspectives and identifies six research gaps.

  5. From Question Answering to Task Completion: A Survey on Agent System and Harness Design

    cs.AI 2026-06 unverdicted novelty 4.0

    Survey framing LLM agents as model-plus-harness systems, decomposing harness responsibilities, mapping them to tasks, and highlighting open challenges in evaluation, safety, and co-evolution.