CL-Bench is the first expert-validated benchmark for continual learning in frontier LLMs across six real-world domains, showing limited gains and that naive in-context learning outperforms dedicated memory systems.
super hub Canonical reference
WebArena: A Realistic Web Environment for Building Autonomous Agents
Canonical reference. 76% of citing Pith papers cite this work as background.
284
Pith papers citing it
27
external citations · Pith
Background
76% of classified citations
abstract
With advances in generative AI, there is now potential for autonomous agents to manage daily tasks via natural language commands. However, current agents are primarily created and tested in simplified synthetic environments, leading to a disconnect with real-world scenarios. In this paper, we build an environment for language-guided agents that is highly realistic and reproducible. Specifically, we focus on agents that perform tasks on the web, and create an environment with fully functional websites from four common domains: e-commerce, social forum discussions, collaborative software development, and content management. Our environment is enriched with tools (e.g., a map) and external knowledge bases (e.g., user manuals) to encourage human-like task-solving. Building upon our environment, we release a set of benchmark tasks focusing on evaluating the functional correctness of task completions. The tasks in our benchmark are diverse, long-horizon, and designed to emulate tasks that humans routinely perform on the internet. We experiment with several baseline agents, integrating recent techniques such as reasoning before acting. The results demonstrate that solving complex tasks is challenging: our best GPT-4-based agent only achieves an end-to-end task success rate of 14.41%, significantly lower than the human performance of 78.24%. These results highlight the need for further development of robust agents, that current state-of-the-art large language models are far from perfect performance in these real-life tasks, and that WebArena can be used to measure such progress.
hub tools
citation-role summary
background 35
dataset 4
baseline 2
method 1
citation-polarity summary
claims ledger
- abstract With advances in generative AI, there is now potential for autonomous agents to manage daily tasks via natural language commands. However, current agents are primarily created and tested in simplified synthetic environments, leading to a disconnect with real-world scenarios. In this paper, we build an environment for language-guided agents that is highly realistic and reproducible. Specifically, we focus on agents that perform tasks on the web, and create an environment with fully functional websites from four common domains: e-commerce, social forum discussions, collaborative software develop
authors
co-cited works
representative citing papers
MedMemoryBench supplies a 2,000-session synthetic medical trajectory dataset and an evaluate-while-constructing streaming protocol to expose memory saturation and reasoning failures in current agent architectures for personalized healthcare.
Agent-BRACE improves LLM agent performance on long-horizon partially observable tasks by 5.3-14.5% through a decoupled belief state of verbalized atomic claims with certainty labels that keeps context length constant.
A new native-runtime benchmark reveals that current frontier AI agents succeed on at most 62 percent of realistic long-horizon CLI tasks.
WindowsWorld benchmark shows leading GUI agents achieve under 21% success on multi-application professional tasks, with failures especially on conditional judgment across three or more apps and inefficient execution.
MCP-Atlas is a new benchmark with 1000 tasks on production MCP servers that uses claim-level scoring to evaluate LLM agents on realistic multi-step tool-use competency.
AgentDojo introduces an extensible evaluation framework populated with realistic agent tasks and security test cases to measure prompt injection robustness in tool-using LLM agents.
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.
A²utoLPBench is a generator that produces unlimited LP word problems with ground-truth answers known by construction via inverse-KKT, bundled with a Docker environment for agent evaluation.
Self-GC governs agent context as indexed objects with planner-proposed actions, achieving 84.85% no-impact on future continuations on a hard set versus 54-70% for baselines.
PrincipalBench exposes a sharp split in frontier LLMs between selective and over-refusing behavior on multi-party loyalty, with prompt scaffolding and KL distillation reducing harm rates but only along an existing leak/over-refusal trade-off.
SpreadsheetBench 2 provides 321 expert-validated tasks from authentic business data showing frontier LLMs reach only 34.89% overall accuracy on end-to-end spreadsheet workflows.
AOI adds keyframe capture, volume-gated audio transcription, and visual narration to computer-use agents, producing +17 to +48 pp gains over screenshot baselines on DynaCU-Bench with no retraining.
SEATauBench is the first agent benchmark for SEA languages, finding that performance holds for language-only changes but degrades sharply with full domain localization.
CFAgentBench is a new reproducible benchmark for construction-finance AI agents featuring 35 mock apps, 1,014 tasks, and a money-movement guard, with initial tests showing pass^1 of 0.67 dropping to pass^5 of 0.38.
Introduces the Power Systems Agent Benchmark with 41 task families across eight power engineering areas for executable evaluation of AI agents using deterministic feasibility checks.
StaminaBench evaluates coding agents over 100 procedurally generated change requests to a REST API, finding that tested models fail within 5-6 turns without feedback but improve up to 12x with test feedback and good harnesses.
SafeClawBench supplies 600 staged adversarial tasks and three separate endpoints that show semantic acceptance, audit evidence, and sandbox-observed harm are distinct failure modes in tool-using LLM agents.
EComAgentBench is a new benchmark with 662 tasks distributing hidden intent across sources and using source-tagged rubrics, where the strongest of seven tested models reaches only 57.1% accuracy.
Attackers can force LLM guardrails into extended reasoning loops via optimized payloads, causing 13-63x token amplification and up to 148x latency in agent systems.
The paper builds SOPBench showing frequent SOP violations in agentic browsers and introduces SOPGuard to enforce the policy with low overhead in BrowserOS.
Introduces a stakeholder-centric benchmark showing current web agents fail all tested prompt injection objectives, with failures falling into stealthy parasitism, misaligned disruption, or compounded failure modes.
Layer-isolated evaluation decomposes LLM agents into per-layer deterministic no-LLM test slices whose locked baselines localize regressions that aggregate pass rates mask.
WebChallenger introduces PageMem and three architecture mechanisms to achieve competitive web navigation with open-weight LLMs on WebArena, VisualWebArena, Online-Mind2Web, and WorkArena without fine-tuning or site adapters.
citing papers explorer
-
Continual Learning Bench: Evaluating Frontier AI Systems in Real-World Stateful Environments
CL-Bench is the first expert-validated benchmark for continual learning in frontier LLMs across six real-world domains, showing limited gains and that naive in-context learning outperforms dedicated memory systems.
-
MedMemoryBench: Benchmarking Agent Memory in Personalized Healthcare
MedMemoryBench supplies a 2,000-session synthetic medical trajectory dataset and an evaluate-while-constructing streaming protocol to expose memory saturation and reasoning failures in current agent architectures for personalized healthcare.
-
Agent-BRACE: Decoupling Beliefs from Actions in Long-Horizon Tasks via Verbalized State Uncertainty
Agent-BRACE improves LLM agent performance on long-horizon partially observable tasks by 5.3-14.5% through a decoupled belief state of verbalized atomic claims with certainty labels that keeps context length constant.
-
WildClawBench: A Benchmark for Real-World, Long-Horizon Agent Evaluation
A new native-runtime benchmark reveals that current frontier AI agents succeed on at most 62 percent of realistic long-horizon CLI tasks.
-
WindowsWorld: A Process-Centric Benchmark of Autonomous GUI Agents in Professional Cross-Application Environments
WindowsWorld benchmark shows leading GUI agents achieve under 21% success on multi-application professional tasks, with failures especially on conditional judgment across three or more apps and inefficient execution.
-
MCP-Atlas: A Large-Scale Benchmark for Tool-Use Competency with Real MCP Servers
MCP-Atlas is a new benchmark with 1000 tasks on production MCP servers that uses claim-level scoring to evaluate LLM agents on realistic multi-step tool-use competency.
-
AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents
AgentDojo introduces an extensible evaluation framework populated with realistic agent tasks and security test cases to measure prompt injection robustness in tool-using LLM agents.
-
OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments
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.
-
A$^{2}$utoLPBench: An Auto-Generated, Agent-Friendly LP Benchmark via Inverse-KKT Construction
A²utoLPBench is a generator that produces unlimited LP word problems with ground-truth answers known by construction via inverse-KKT, bundled with a Docker environment for agent evaluation.
-
Self-GC: Self-Governing Context for Long-Horizon LLM Agents
Self-GC governs agent context as indexed objects with planner-proposed actions, achieving 84.85% no-impact on future continuations on a hard set versus 54-70% for baselines.
-
Whose Side Is Your Agent On? Multi-Party Principal Loyalty in LLM Agents
PrincipalBench exposes a sharp split in frontier LLMs between selective and over-refusing behavior on multi-party loyalty, with prompt scaffolding and KL distillation reducing harm rates but only along an existing leak/over-refusal trade-off.
-
SpreadsheetBench 2: Evaluating Agents on End-to-End Business Spreadsheet Workflows
SpreadsheetBench 2 provides 321 expert-validated tasks from authentic business data showing frontier LLMs reach only 34.89% overall accuracy on end-to-end spreadsheet workflows.
-
Agent-Computer Observation Interfaces Enable Dynamic Computer Use
AOI adds keyframe capture, volume-gated audio transcription, and visual narration to computer-use agents, producing +17 to +48 pp gains over screenshot baselines on DynaCU-Bench with no retraining.
-
SEATauBench: Adapting Tool-Agent-User Evaluation Into Low-Resource Southeast Asian Languages
SEATauBench is the first agent benchmark for SEA languages, finding that performance holds for language-only changes but degrades sharply with full domain localization.
-
CFAgentBench: A Reproducible Environment and Benchmark for Autonomous Construction-Finance Agents
CFAgentBench is a new reproducible benchmark for construction-finance AI agents featuring 35 mock apps, 1,014 tasks, and a money-movement guard, with initial tests showing pass^1 of 0.67 dropping to pass^5 of 0.38.
-
Power Systems Agent Benchmark: Executable Evaluation of AI Agents in Electric Power Engineering
Introduces the Power Systems Agent Benchmark with 41 task families across eight power engineering areas for executable evaluation of AI agents using deterministic feasibility checks.
-
StaminaBench: Stress-Testing Coding Agents over 100 Interaction Turns
StaminaBench evaluates coding agents over 100 procedurally generated change requests to a REST API, finding that tested models fail within 5-6 turns without feedback but improve up to 12x with test feedback and good harnesses.
-
SafeClawBench: Separating Semantic, Audit-Evidence, and Sandbox Harm in Tool-Using LLM Agents
SafeClawBench supplies 600 staged adversarial tasks and three separate endpoints that show semantic acceptance, audit evidence, and sandbox-observed harm are distinct failure modes in tool-using LLM agents.
-
EComAgentBench: Benchmarking Shopping Agents on Long-Horizon Tasks with Distributed Hidden Intent
EComAgentBench is a new benchmark with 662 tasks distributing hidden intent across sources and using source-tagged rubrics, where the strongest of seven tested models reaches only 57.1% accuracy.
-
From Shield to Target: Denial-of-Service Attacks on LLM-Based Agent Guardrails
Attackers can force LLM guardrails into extended reasoning loops via optimized payloads, causing 13-63x token amplification and up to 148x latency in agent systems.
-
Same-Origin Policy for Agentic Browsers
The paper builds SOPBench showing frequent SOP violations in agentic browsers and introduces SOPGuard to enforce the policy with low overhead in BrowserOS.
-
Who Pays the Price? Stakeholder-Centric Prompt Injection Benchmarking for Real-world Web Agents
Introduces a stakeholder-centric benchmark showing current web agents fail all tested prompt injection objectives, with failures falling into stealthy parasitism, misaligned disruption, or compounded failure modes.
-
Layer-Isolated Evaluation: Gating the Deterministic Scaffold of a Production LLM Agent with a No-LLM, Regression-Locked Test Harness
Layer-isolated evaluation decomposes LLM agents into per-layer deterministic no-LLM test slices whose locked baselines localize regressions that aggregate pass rates mask.
-
WebChallenger: A Reliable and Efficient Generalist Web Agent
WebChallenger introduces PageMem and three architecture mechanisms to achieve competitive web navigation with open-weight LLMs on WebArena, VisualWebArena, Online-Mind2Web, and WorkArena without fine-tuning or site adapters.
-
Bittensor Agent Arenas as a Trajectory Primitive: Distilling a Shopping Agent from ShoppingBench Subnet Traces
Trajectories from a Bittensor ShoppingBench subnet arena, filtered to retain only agentic tool-calling behavior, enable SFT+GRPO post-training of Qwen3-4B to 42.7% ASR on leak-guarded held-out tests, nearly matching synthetic-data baselines with a fraction of a day's data.
-
Beyond Goodhart's Law: A Dynamic Benchmark for Evaluating Compliance in Multi-Agent Systems
MAC-Bench is a new adversarial benchmark that converts legal texts into executable scenarios via the SERV pipeline to measure procedural compliance in multi-agent LLM systems using CSR and MG metrics.
-
Humans' ALMANAC: A Human Collaboration Dataset of Action-Level Mental Model Annotations for Agent Collaboration
ALMANAC is a new dataset of 2,987 annotated dyadic collaboration actions from the Map Task, each with theory-informed mental model annotations for self-reasoning, partner intent, and team goal, used to benchmark six LLMs on predicting next-turn behavior and mental models.
-
Agent Planning Benchmark: A Diagnostic Framework for Planning Capabilities in LLM Agents
Introduces APB benchmark with 4209 cases across 22 domains to diagnose planning in 12 MLLMs and shows it improves downstream execution when used for refinement.
-
Domain-Conditioned Safety in Frontier Computer-Using Agents: A 793-Episode Browser Benchmark, a Coding-Domain Cross-Reference, and a Reproducibility Audit of Recent Red-Teaming
Frontier browser agents show strong resistance to hand-crafted multi-step prompt injections (0/140 success), unlike coding agents (up to 100%), indicating domain-conditioned safety and that prior high ASR reports may not generalize.
-
HLL: Can Agents Cross Humanity's Last Line of Verification?
HLL is a new benchmark that evaluates eight frontier multimodal agents on closed-loop interactive CAPTCHA solving, showing sharp performance drops under realism stressors and trace validation.
-
OR-Space: A Full-Lifecycle Workspace Benchmark for Industrial Optimization Agents
OR-Space is a benchmark for LLM agents performing full-lifecycle optimization tasks across Build, Revise, and Explain modes in executable multi-artifact workspaces.
-
AndroidDaily: A Verifiable Benchmark for Mobile GUI Agents on Real-World Closed-Source Applications
AndroidDaily supplies 350 verifiable tasks on 94 closed-source Android apps evaluated by GRADE (87.37% human agreement), with the strongest model achieving 62% success.
-
AgenticVBench: Can AI Agents Complete Real-World Post-Production Tasks?
AgenticVBench evaluates frontier VLMs on 100 real-world video post-production tasks across four families, with the best agent stack scoring just over 30% versus human experts.
-
LogDx-CI: Benchmarking Log Reduction Tools for LLM Root-Cause Diagnosis
LogDx-CI benchmark shows hybrid grep+tail reducers achieve top diagnosis quality at low cost, agent loops shrink quality variance across reducers, and cross-family LLM summarizers outperform same-family pairs.
-
JobBench: Aligning Agent Work With Human Will
JobBench is a new benchmark with 130 occupational tasks where the best of 36 tested AI models achieves only 45.9% success.
-
CyberEvolver: Structured Self-Evolution for Cybersecurity Agents On the Fly
CyberEvolver introduces a four-layer self-evolving agent architecture with trace-to-diagnosis and population beam search that raises seed agent success rates by 13.6% on CTF, exploitation, and penetration tasks across four LLMs.
-
ScaleWoB: Guiding GUI Agents with Coding Agents via Large-Scale Environmental Synthesis
ScaleWoB generates 100+ synthetic interactive GUI environments and 1000+ verifiable tasks as web pages, releasing a 120-task mobile benchmark where state-of-the-art agents achieve 27.92% success (17.82% on long-horizon tasks) versus 92.08% for humans, with synthetic results generalizing to real apps
-
VISTA: An End-to-End Benchmark for Visual Spec-to-Web-App Coding Agents
VISTA is a new benchmark for end-to-end visual spec-to-web-app generation by LLM agents, featuring five prompt conditions, manual UI annotations, multi-metric evaluation, and results on four agent systems showing partial decoupling of visual and functional performance.
-
Boiling the Frog: A Multi-Turn Benchmark for Agentic Safety
Boiling the Frog is a new stateful multi-turn benchmark that finds an aggregate 44.4% strict attack success rate for incremental safety violations across nine AI models, with rates ranging from 20.5% to 92.9%.
-
IdleSpec: Exploiting Idle Time via Speculative Planning for LLM Agents
IdleSpec improves LLM agent accuracy by generating and aggregating speculative plans during idle time between tool calls and observations using complementary drafting strategies.
-
MemGym: a Long-Horizon Memory Environment for LLM Agents
MemGym unifies agent gyms into a memory benchmark with isolated scoring across tool-use, research, coding, and computer-use regimes plus a lightweight reward model for tractable coding evaluation.
-
Llamas on the Web: Memory-Efficient, Performance-Portable, and Multi-Precision LLM Inference with WebGPU
LlamaWeb is a WebGPU backend for llama.cpp that uses static memory planning, tunable kernels, and templated multi-precision support to cut memory use by 29-33% and raise decode throughput by 45-69% versus prior browser frameworks on tested hardware.
-
Agent Meltdowns: The Road to Hell Is Paved with Helpful Agents
The paper defines accidental meltdowns as unsafe agent behavior triggered by benign errors and reports that such meltdowns occur in 64.7% of evaluated rollouts across GPT, Grok, and Gemini agents.
-
SCICONVBENCH: Benchmarking LLMs on Multi-Turn Clarification for Task Formulation in Computational Science
SCICONVBENCH is a new benchmark evaluating LLMs on multi-turn disambiguation and inconsistency resolution for task formulation in computational science, with frontier models reaching only 52.7% success on fluid mechanics disambiguation cases.
-
Skim: Speculative Execution for Fast and Efficient Web Agents
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.
-
Known By Their Actions: Fingerprinting LLM Browser Agents via UI Traces
UI traces of actions and timings from LLM browser agents enable identification of the underlying model with up to 96% F1 across 14 models and multiple tasks.
-
$\pi$-Bench: Evaluating Proactive Personal Assistant Agents in Long-Horizon Workflows
π-Bench is a new benchmark for evaluating proactive personal assistant agents on 100 multi-turn tasks that include hidden intents, inter-task dependencies, and cross-session continuity.
-
Beyond Binary: Reframing GUI Critique as Continuous Semantic Alignment
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.
-
ClawForge: Generating Executable Interactive Benchmarks for Command-Line Agents
ClawForge is a generator framework that creates reproducible executable benchmarks for command-line agents under state conflict, with ClawForge-Bench showing frontier models reach at most 45.3% strict accuracy and that state inspection drives most performance gaps.
-
MMSkills: Towards Multimodal Skills for General Visual Agents
MMSkills packages multimodal procedural knowledge into state-conditioned skills with text, state cards, and multi-view keyframes, generated from public trajectories via an agentic process and used at inference via branch-loaded inspection to improve visual agents on GUI and game benchmarks.