OSWorld provides the first unified real-computer benchmark for open-ended multimodal agent tasks, exposing large performance gaps between humans and state-of-the-art LLM/VLM agents.
hub Mixed citations
GPT-4V(ision) is a Generalist Web Agent, if Grounded
Mixed citation behavior. Most common role is background (67%).
32
Pith papers citing it
Background
67% of classified citations
abstract
The recent development on large multimodal models (LMMs), especially GPT-4V(ision) and Gemini, has been quickly expanding the capability boundaries of multimodal models beyond traditional tasks like image captioning and visual question answering. In this work, we explore the potential of LMMs like GPT-4V as a generalist web agent that can follow natural language instructions to complete tasks on any given website. We propose SEEACT, a generalist web agent that harnesses the power of LMMs for integrated visual understanding and acting on the web. We evaluate on the recent MIND2WEB benchmark. In addition to standard offline evaluation on cached websites, we enable a new online evaluation setting by developing a tool that allows running web agents on live websites. We show that GPT-4V presents a great potential for web agents -- it can successfully complete 51.1 of the tasks on live websites if we manually ground its textual plans into actions on the websites. This substantially outperforms text-only LLMs like GPT-4 or smaller models (FLAN-T5 and BLIP-2) specifically fine-tuned for web agents. However, grounding still remains a major challenge. Existing LMM grounding strategies like set-of-mark prompting turns out to be not effective for web agents, and the best grounding strategy we develop in this paper leverages both the HTML structure and visuals. Yet, there is still a substantial gap with oracle grounding, leaving ample room for further improvement. All code, data, and evaluation tools are available at https://github.com/OSU-NLP-Group/SeeAct.
hub tools
citation-role summary
background 7
baseline 1
dataset 1
citation-polarity summary
representative citing papers
Skim profiles website patterns offline to enable fast-path speculative execution for web agents, cutting median cost by 1.9x and latency by 33.4% with no accuracy loss on benchmarks.
BBCritic reframes GUI critique as continuous semantic alignment via contrastive learning in an affordance space, outperforming larger binary SOTA models on a new four-level hierarchical benchmark without extra annotations.
SDP constructs a task-induced state space from raw text by having agents commit to and certify natural-language predicates as states, enabling structured planning and analysis in unstructured language environments.
Presents CUActSpot benchmark and renderer-LLM data synthesis that lets a 4B model outperform larger open-source models on complex computer interactions.
Weblica scales RL training for visual web agents by building thousands of reproducible environments through HTTP caching for stable replays and LLM synthesis from real sites, yielding an 8B model that beats similar open baselines on navigation benchmarks.
Agentic browsers are vulnerable to 20 web and LLM attacks with 18 implemented, exposing five failure modes across four major LLM models that require redesign before safe deployment.
VLMs detect primitive motion in UI animations reliably but show inconsistent high-level interpretation of purposes and meanings, with large gaps relative to human performance.
UIPress is the first encoder-side learned optical compression method for UI-to-Code that compresses visual tokens to 256, outperforming the uncompressed baseline by 7.5% CLIP score and the best inference-time baseline by 4.6% while delivering 9.1x TTFT speedup.
ClawBench is a benchmark of 153 live-web tasks where AI agents achieve low success rates, e.g. 33.3% for Claude Sonnet 4.6.
GUIDE decomposes GUI agent evaluation into trajectory segmentation, subtask diagnosis, and overall summary to deliver higher accuracy and structured error reports than holistic baselines.
OmniGUI is the first step-level benchmark supplying interleaved image, audio, and video inputs across 709 expert episodes in 29 smartphone apps to evaluate multimodal GUI agents.
GiGPO adds a hierarchical grouping mechanism to group-based RL so that LLM agents receive both global trajectory and local step-level credit signals, yielding >12% gains on ALFWorld and >9% on WebShop over GRPO while keeping the same rollout and memory footprint.
MementoGUI introduces a modular memory-control framework with working and episodic memory operators that improves long-horizon GUI agent performance over history-replay and text-only baselines.
Web agents should default to planning a complete task program before observing live web content to reduce prompt injection exposure, since WebArena tasks are compatible and 80% need no runtime LLM calls.
PageGuide grounds LLM answers in webpage DOM elements using visual overlays for find, guide, and hide modes, yielding measurable gains in a 94-user study.
VLAA-GUI adds mandatory visual verifiers, multi-tier loop breakers, and on-demand search to GUI agents, reaching 77.5% on OSWorld and 61.0% on WindowsAgentArena with some models exceeding human performance.
Closed-loop VLM agents using multi-view reasoning, object-centered visualization, and single-axis rotation prediction achieve superior text-guided 6D pose rearrangement for target objects in scenes.
WebChain supplies the largest open dataset of real human web trajectories with triple-modal alignment and a dual mid-training method that separates grounding from planning to improve web agents.
The paper proposes Amazon-Bench, a functionality-grounded benchmark for web agents in e-commerce that generates diverse task queries from webpage elements and evaluates both task performance and safety risks.
Introduces an app-content instrumentation framework and benchmark showing that examined GUI agents suffer 42.0% and 36.1% average misleading rates from third-party content in dynamic and static tests respectively.
Introduces secondary risks as a new class of LLM failures from benign prompts, defines two primitives, proposes SecLens search framework, and releases SecRiskBench showing risks are widespread across 16 models.
VS-Bench is a new benchmark of ten visual multi-agent environments that measures VLMs on element recognition, next-action prediction, and normalized episode return, showing strong perception but large gaps in reasoning and decision-making with the best model at 46.6% prediction accuracy and 31.4% of
WebCanvas creates a dynamic benchmark for web agents with a noise-resistant evaluation metric, the Mind2Web-Live dataset of 542 tasks, and open-source tools and agent framework for ongoing online testing.
citing papers explorer
-
Group-in-Group Policy Optimization for LLM Agent Training
GiGPO adds a hierarchical grouping mechanism to group-based RL so that LLM agents receive both global trajectory and local step-level credit signals, yielding >12% gains on ALFWorld and >9% on WebShop over GRPO while keeping the same rollout and memory footprint.
-
A Functionality-Grounded Benchmark for Evaluating Web Agents in E-commerce Domains
The paper proposes Amazon-Bench, a functionality-grounded benchmark for web agents in e-commerce that generates diverse task queries from webpage elements and evaluates both task performance and safety risks.
-
Mobile GUI Agents under Real-world Threats: Are We There Yet?
Introduces an app-content instrumentation framework and benchmark showing that examined GUI agents suffer 42.0% and 36.1% average misleading rates from third-party content in dynamic and static tests respectively.
-
Exploring the Secondary Risks of Large Language Models
Introduces secondary risks as a new class of LLM failures from benign prompts, defines two primitives, proposes SecLens search framework, and releases SecRiskBench showing risks are widespread across 16 models.
-
VS-Bench: Evaluating VLMs for Strategic Abilities in Multi-Agent Environments
VS-Bench is a new benchmark of ten visual multi-agent environments that measures VLMs on element recognition, next-action prediction, and normalized episode return, showing strong perception but large gaps in reasoning and decision-making with the best model at 46.6% prediction accuracy and 31.4% of