Tool use, function calling, orchestration and the protocols that let models act rather than only answer.
arXiv:2609.08149v1 Announce Type: new Abstract: SWE-Bench Pro has emerged as a standard benchmark for evaluating software engineering agents on challenging repository-level tasks. However, our analys...
The paper introduces ARC, a lightweight hierarchical policy that learns to configure LLM‑based agent systems on a per‑query basis by treating each configuration as a temporally extended option in a semi‑Markov decision process. Unlike fixed templates or hand‑tuned heuristics, ARC dynamically selects workflows, tools, token budgets, and prompts tailored to the difficulty of each query. Experiments on reasoning, tool‑use, and agentic benchmarks show that ARC outperforms budget‑matched tool‑augmented LLMs, boosting reasoning accuracy by 31.3%, tool‑use accuracy by 13.95%, and doubling success on the τ‑Bench Airline Pass task from 9.0% to 18.0%.
FrogNano is a 4B coding agent trained exclusively with reinforcement learning on about 1,500 synthetic software engineering environments. Its training leverages an online task synthesis pipeline that generates tasks at the current agent’s learnability frontier, improving performance without distilling from larger models. The report details the methodology, evaluates the agent across diverse environments, and analyzes its effectiveness as a lightweight coding agent for minimal hardware.
UnitBoost proposes a non‑generative merge operator to manage compound LLM systems, replacing the opaque higher‑level LLM that traditionally coordinates worker outputs. The operator maps worker outputs to slot‑value proposals, uses a constrained argmax to assemble the final answer, and explicitly tracks unfilled slots as residuals for subsequent rounds, thereby achieving order‑free processing and unit provenance. Across three held‑out benchmarks, UnitBoost outperforms both gold‑label‑selected candidates and input‑matched generative managers, improving compound‑system performance by up to 0.182 points and raising FanOutQA cell F1 from 0.4778 to 0.5524.
arXiv:2609.05901v1 Announce Type: cross Abstract: Autonomous LLM agents can turn untrusted content into effectful actions such as payments and permission changes. If the same process interprets this...
arXiv:2609.08062v1 Announce Type: new Abstract: Tool-using language agents can delegate and revoke permissions while acting through external services. We show that two authorization histories can hav...
The paper introduces Online Surrogate Repair (OSR), a closed‑loop algorithm that decouples the frequency of high‑fidelity evaluations from the length of an agent’s search by selectively updating a surrogate model with sparse, high‑fidelity data. An acquisition rule determines which candidate designs receive expensive evaluations, and the resulting labels refine the surrogate for subsequent episodes. Experiments on synthetic environments and the MADE benchmark show that OSR can reduce regret more efficiently than fixed‑surrogate approaches, requiring fewer oracle queries than high‑fidelity feedback after every episode.
X2Streaming-ASR introduces a method for streaming automatic speech recognition that separates the decision of when to commit a transcript from what to commit. The approach uses a three‑stage training process: first establishing streaming capability, then warm‑starting a commit policy with automatically probed trajectories, and finally refining the policy with character‑level, segment‑assigned group‑relative rewards for accuracy and latency. On AISHELL‑1/2/3 and WenetSpeech datasets, the system achieves mean character‑level commit latencies of 27–84 ms, far lower than baseline systems, while also attaining the best streaming character error rates on AISHELL‑1 and AISHELL‑3.
The paper introduces Open Tabular Insight Extraction (OpenTI), a unified framework aimed at democratizing access to insights from large table corpora. It highlights how current research is fragmented across domains like table QA, text‑to‑SQL, and data analysis agents, and shows that existing systems and benchmarks fall short of covering the full end‑to‑end scope of OpenTI. The authors propose a consolidated terminology, conduct a systematic review, and outline a research agenda for developing comprehensive OpenTI systems, evaluation methods, and interaction paradigms.
The paper introduces the Multi-Session Personalized Tool Calling (MPT) benchmark, containing 4,695 instances across 459 multi‑session histories that test Preference Recall, Induction, and Transfer. It proposes PRefine, a test‑time memory method that refines a user’s latent preference via a generate‑verify‑refine loop. Experiments with five LLMs show that PRefine outperforms existing memory systems and even full‑history prompting on Preference Transfer, suggesting that personalized agents should encode behavior as preferences rather than merely storing past interactions.
AI agents increasingly act through tools and delegated authority, but general incident repositories rarely capture the mechanisms needed to compare public failures with agent-security evaluations. We...
Studies of the Model Context Protocol (MCP) server ecosystem draw their samples in ways that quietly select for servers that work: reference sets, popularity lists, hand-curated frames, or pipelines t...
Build and launch cloud agents with the Agents API, a managed service powered by the Codex harness for orchestration, long-running sessions, and tool use.
Simon Willison created a .blend URL Viewer tool that lets users view a Blender model of a Fabergé egg themed after the TV show Pluribus directly in their browser. He generated the egg image using ChatGPT Images 2.5, then fed it to GPT‑6 Astra with a custom Blender skill to produce several .blend files. The viewer, built with JavaScript, is now part of his tools collection for easy access to the resulting 3D model.
A research idea may be novel, coherent, and scientifically plausible, yet its proposed method may remain insufficiently specified for faithful implementation. We study the codification readiness of im...
Persistent language agents need stored experience to remain available across time, while each answer requires evidence suited to a particular question. A superseded fact can mislead a current-state an...
To address insufficient contextualization, weak generalization, and poor scenario adaptation in tourism meteorological services, we propose SmartWeatherAgent--a unified three-stage architecture integr...
Large language models (LLMs) have become ubiquitous in software development, with LLM-based automated program repair tools increasingly used during code review. In this report, we explore the iterativ...
Compute governance today is a governance of training: the thresholds, reporting requirements, and frontier-AI regimes now in force attach to training compute and treat the trained model as the regulat...
Mistral AI assisted a European energy operator in modernizing a substantial legacy codebase by migrating 40,000 lines of Fortran 77 to C++. The project showcases the practical application of AI agents in refactoring complex code, detailing the process and key lessons learned. It serves as a case study for similar modernization efforts in the industry.