REVIEW 23 cited by
Is Your LLM Secretly a World Model of the Internet? Model-Based Planning for Web Agents
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
Language agents based on large language models (LLMs) have demonstrated great promise in automating web-based tasks. Recent work has shown that incorporating advanced planning algorithms, e.g., tree search, is advantageous over reactive planning for web agents. However, unlike simulated sandbox environments, real-world environments such as the web are rife with irreversible actions. This undermines the feasibility of backtracking, a cornerstone of (tree) search. Overly relying on test-time search also hurts efficiency. We advocate model-based planning for web agents that employs a world model to simulate and deliberate over the outcome of each candidate action before committing to one. We systematically explore this paradigm by (1) Proposing a model-based planning framework, WebDreamer, which employs LLMs to serve as both world models and value functions; (2) Training specialized LLMs as world models with a scalable data synthesis pipeline. Empirical results demonstrate that WebDreamer achieves substantial performance improvements over reactive baselines. It is competitive, while being 4-5 times more efficient, with tree search in sandbox environments (VisualWebArena) and also works effectively on real-world websites (Online-Mind2Web and Mind2Web-Live). Furthermore, our trained world model, Dreamer-7B, performs comparable to GPT-4o, highlighting the potential of specialized world models for efficient and effective planning in complex web environments.
Forward citations
Cited by 23 Pith papers
-
AppDeltaWorld: Transition-Grounded Delta Code World Model for Mobile GUI Agents
AppDeltaWorld predicts mobile GUI transitions as code updates retrieved under action constraints, and its generated trajectories improve an 8B mobile agent on several benchmarks.
-
RMSWeb: Reflection, Failure-Mode Mining, and Salvage-DS for Web Agent Reinforcement Learning
RMSWeb combines reflection-conditioned collection, failure-mode mining, and a competence-gated offline RL objective to improve compact web agents' success and step efficiency on live-web benchmarks.
-
Why Are GUI Agents Correct but Late? Decode on the Decision-Time Critical Path, Tested with Pre-Compiled Policy Trees
Removing autoregressive decode from the decision-time critical path via pre-compiled guarded policy trees recovers contested GUI action windows when outcomes are enumerable in advance.
-
Agent Learning via Early Experience
Training agents to predict consequences of their own alternative actions (implicit world modeling) or to reflect on why expert actions were better (self-reflection) consistently improves task success, OOD generalizati...
-
SSRL: Self-Search Reinforcement Learning
SSRL, a training pipeline that uses an LLM's own repeated sampling as a search environment for RL, improves question answering without external tools and transfers to real search engines.
-
OS-MAP: How Far Can Computer-Using Agents Go in Breadth and Depth?
A new 416-task desktop benchmark with a five-level automation taxonomy and a demand-based generalization hierarchy shows current computer-using agents succeed on roughly one in nine tasks, with near-zero success on hi...
-
WebSynthesis: World-Model-Guided MCTS for Efficient WebUI-Trajectory Synthesis
A world-model-guided MCTS pipeline synthesizes 4k web navigation trajectories and yields a WebArena Pass@3 success rate of 20.15%, above OS-Genesis (18.66%) and AgentTrek (11.94%).
-
Universal Retrieval for Multimodal Trajectory Modeling
Introduces multimodal trajectory retrieval, constructs the GAE-Bench benchmark from five GUI agent datasets, and trains GAE-Retriever, which outperforms zero-shot retrieval baselines.
-
Build the web for agents, not agents for the web
The paper proposes a paradigm shift: design a standardized Agentic Web Interface for AI agents, rather than adapting agents to human-facing websites.
-
GOBench: Benchmarking Geometric Optics Generation and Understanding of MLLMs
GOBench measures how well multimodal AI models generate and understand geometric optics, finding that even top models make frequent physical errors.
-
Improving Large Language Model Planning with Action Sequence Similarity
Selecting in-context examples by action-sequence similarity to a model-generated draft plan improves LLM planning accuracy over random or problem-similarity selection.
-
ScaleTrack: Scaling and back-tracking Automated GUI Agents
ScaleTrack combines multiple GUI grounding datasets and a back-tracking training objective, predicting the previous action from the current screenshot, to improve GUI agent planning, achieving modest gains on AndroidC...
-
WebLists: Extracting Structured Information From Complex Interactive Websites Using Executable LLM Agents
BardeenAgent records a single extraction on a webpage, generalizes it with CSS selectors, and replays it across all matching items, reaching 66% recall on a new 200-task WebLists benchmark versus 31% for the best baseline.
-
Trajectory World Models for Heterogeneous Environments
A pre-trained trajectory world model with interleaved temporal and variate attention achieves positive transfer across heterogeneous control environments, improving transition prediction, off-policy evaluation, and mo...
-
Software Engineering for and with GUI Agent
A survey of 336 GUI-agent papers finds rapid growth alongside weak engineering support for recovery, human oversight, maintainability, and privacy, and calls for lifecycle-centered testing and governance.
-
TAPO: Transition-Aware Policy Optimization for LLM Agents
Interleaving action-conditioned next-observation supervision with group RL on a shared LLM backbone consistently lifts long-horizon agent success over pure policy optimization.
-
Branch-and-Browse: Efficient and Controllable Web Exploration with Tree-Structured Reasoning and Action Memory
Branch-and-Browse, a tree-structured web agent with page-level action memory, reports 35.8% success and up to 40.4% less time than the cited Tree Search baseline on WebArena.
-
Pipelined Decoder for Efficient Context-Aware Text Generation
The pipelined decoder trains a T5 model to generate multiple subsequences in parallel with partial token dependencies, achieving 1.7x to 7x faster inference on context-aware generation tasks with roughly unchanged quality.
-
Agent-RewardBench: Towards a Unified Benchmark for Reward Modeling across Perception, Planning, and Safety in Real-World Multimodal Agents
A new step-level reward-modeling benchmark for multimodal agents shows current MLLMs reach at most 61.6 percent accuracy, and benchmark score correlates strongly (r=0.981 across five models) with downstream A* search ...
-
R2D2: Remembering, Replaying and Dynamic Decision Making with a Reflective Agentic Memory
R2D2 combines a replay-buffer graph with A* search and reflective memory to boost web-agent navigation and task completion on WebArena.
-
The BrowserGym Ecosystem for Web Agent Research
A unified benchmarking ecosystem for web agents, plus a large cross-benchmark evaluation of six LLMs in which Claude-3.5-Sonnet is the strongest overall.
-
Foundation Models for CPS-IoT: Opportunities and Challenges
Current foundation models fall short on CPS-IoT needs in resource efficiency, spatial generalization, long-term context, and knowledge integration; the paper proposes desiderata and a community roadmap.
-
Large Language Models for Planning: A Comprehensive and Systematic Survey
A structured survey of LLM planning methods, benchmarks, and interpretability work, organized around a three-way taxonomy.
Discussion (0). Continue with ORCID to comment.