ASTRA-bench is a benchmark designed to evaluate AI agents’ reasoning and action planning in the presence of evolving personal context and an interactive toolbox. It generates 2,413 scenarios across four protagonists, incorporating longitudinal life events and varying levels of referential, functional, and informational complexity. Evaluation of leading models shows a marked drop in performance under high-complexity conditions, highlighting argument generation as a key bottleneck and exposing limitations in grounding reasoning within messy personal data.
arXiv:2606. 11070v1 Announce Type: cross Abstract: Recent advances in reasoning and tool-calling capabilities of large language models (LLMs) have enabled increasingly capable agentic systems.
AnyAct introduces a universal action layer that consolidates diverse tool capabilities into a self‑evolving action space for AI agents operating in open‑world environments. It tackles the scale dilemma, tool non‑stationarity, and heterogeneous feedback by using hierarchical progressive retrieval and test‑time reliability evolution, while a heterogeneous observation grounding module unifies multi‑modal feedback. Evaluations on LiveMCPBench and the newly created OSMCP benchmark show state‑of‑the‑art performance, with significant gains in task success rate and reduced execution steps, especially for models with limited native capabilities.
arXiv:2606. 13003v1 Announce Type: new Abstract: Prevailing wisdom posits that Multi-Agent Systems (MAS) are superior to Single-Agent Systems (SAS), citing advantages like context protection, parallel processing and distributed decision-making.
arXiv:2607. 25369v1 Announce Type: new Abstract: Agentic systems have rapidly advanced in their ability to interact with real-world environments, leverage external tools, and provide services for users.
ContextPilot is a proactive context‑management framework designed to improve long‑horizon agentic reasoning with large language models. It expands the toolset to include planning, long‑term memory, and soft context offloading, and introduces a reinforcement‑learning strategy that focuses on critical editing decisions and assigns action‑level advantages. Experiments on long‑context QA and deep search tasks demonstrate that ContextPilot achieves stronger performance with a more compact working context, outperforming existing baselines across various base models and benchmarks.
AlphaDiana is a measurement system designed to evaluate agents in verifiable reasoning tasks by providing modular interfaces for datasets, agents, environments, and scorers. It maintains harness-native control loops while separately managing file reuse and memory retention, enabling both macro-level analyses of entire agents and micro-level interventions on components such as tools, skills, and memory. Experiments on open-source models and harnesses across multiple benchmarks reveal that harnesses and component interventions do not consistently improve reasoning accuracy, and that token usage metrics must be interpreted alongside correctness.
arXiv:2605. 07339v2 Announce Type: replace Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities in orchestrating tools for reasoning tasks.
The paper introduces GMA, a new benchmark for evaluating general mobile assistants in realistic, challenging scenarios. GMA expands on existing benchmarks by offering seven open‑source applications across diverse domains and 300 tasks organized into four difficulty tiers, ranging from simple actions to complex multi‑step workflows. The authors evaluate eight state‑of‑the‑art models, showing that performance drops sharply with task complexity, and conduct ablation studies on harness design—such as context retention and state tracking—to demonstrate how these choices can improve outcomes, especially for demanding workflows.
arXiv:2606. 06462v1 Announce Type: new Abstract: Benchmarks are fundamental for evaluating and advancing LLMs and MLLMs by providing standardized and explicit measures of performance.
Agentick is a unified benchmark for sequential decision‑making agents that evaluates RL, LLM, VLM, hybrid, and human agents on 37 procedurally generated tasks across six capability categories, four difficulty levels, and five observation modalities via a single Gymnasium‑compatible interface. It includes a Coding API, oracle reference policies, pre‑built SFT datasets, a composable agent harness, and a live leaderboard. An evaluation of 27 configurations and over 90,000 episodes shows no single approach dominates, with GPT‑5 mini leading overall, PPO excelling in planning and multi‑agent tasks, and the reasoning harness boosting LLM performance by 3‑10×, while ASCII observations outperform natural language.
arXiv:2509. 23071v2 Announce Type: replace-cross Abstract: Retrieval-augmented generation (RAG) agent development is hindered by the lack of executable ground-truth agent-environment interaction trajectories.
Agent Seer is a pipeline that automatically synthesizes realistic evaluation scenarios for AI agents that use external tools, using only the tool’s specification (function names, natural‑language descriptions, and typed parameter schemas). Starting from a single Model Context Protocol (MCP) specification, it enriches raw schemas, generates graded scenarios with synthetic tool outputs, and expands them into mock‑data‑grounded multi‑turn dialogues that demonstrate strong tool‑calling correctness and conversational coherence. Across seven diverse MCP specifications, the pipeline achieves high quality, with parameter‑schema complexity emerging as the main driver of quality variation and argument‑value accuracy identified as the dominant failure mode.