Tool use, function calling, orchestration and the protocols that let models act rather than only answer.
arXiv:2608. 19729v1 Announce Type: new Abstract: Vision-language-model-based embodied agents can complete instructed tasks but often violate safety constraints in the process, a problem recently framed as interactive safety.
arXiv:2608. 19857v1 Announce Type: new Abstract: For AI agents to be useful beyond simple chat, they must hold sensitive user context such as calendars, credentials, health records, and financial data.
arXiv:2606. 15910v2 Announce Type: replace Abstract: A vision-language model can answer a question about a chest radiograph or a pathology slide fluently and confidently while barely using the image, relying instead on language priors.
arXiv:2608. 20106v1 Announce Type: new Abstract: We introduce OenoBench, a wine-domain knowledge benchmark of 3,266 multiple-choice questions across six pillars (regions, grape varieties, viticulture, winemaking, producers, business) and four difficulty tiers.
arXiv:2608. 19380v1 Announce Type: new Abstract: While modern autonomous driving systems excel at perception tasks such as object detection and trajectory prediction, they lack the high-level causal reasoning required to interpret traffic accidents.
arXiv:2608. 19993v1 Announce Type: new Abstract: Loading reusable skill documents into a bounded context window is now the primary way large language model (LLM) agents acquire task-specific capabilities, which makes skill selection a first-order determinant of task performance and token cost.
arXiv:2605. 10828v2 Announce Type: replace Abstract: As large language models are increasingly deployed in retrieval-augmented generation and agentic systems that accumulate extensive context, understanding how distracting information affects long-context performance becomes critical.
DreamBench-SWE is a multi-session benchmark for software-agent memory hygiene in which later software tasks depend on non-inferable evidence from earlier sessions and are scored by executable hidden o...
The standard objection to full automation is demand-side: if humans earn nothing, who buys the output? This confuses an accounting role with a biological species.
The paper discusses a third paradigm shift in software development, termed Software 3.0, where context and reasoning drive behavior. It proposes that Software 3.0 converges to three core components: a generalized database for all persistent state, a large model that performs reasoning and generation, and an agent that orchestrates the interaction between the two. The authors formalize this convergence, present a minimal reference architecture, and analyze its applicability and limits, noting that it applies best to task domains that are expressible, verifiable, externally stateful, and tool-complete.
The study investigates how autonomous coding agents interact with technical documentation, analyzing 557 coding sessions and 33,097 pull requests. Findings reveal that agents primarily engage with agent-facing artefacts, show weak links between documentation consultation and code editing, lack explicit validation sequences, and tend to consult documentation after code changes. The authors propose a two‑lobed cycle model of agent‑documentation interaction and challenge assumptions about actionability and verifiability of agent‑friendly documentation.
Research: A shot-scraper-style JSON API on Bun 1. 4's new Bun.
LLM-based Multi-Agent Systems (MAS) achieve strong performance on complex reasoning tasks by coordinating multiple agents, but at the cost of substantial token consumption. Recent work on automatic topology design, ARG-Designer, has reframed this problem as autoregressive graph generation.
The retrieval layer that helps AI systems navigate, read, and verify information inside even the most complex documents
LLM agents learn by interacting with environments, yet these environments are hand-built and static: blind to an agent's weaknesses, and quickly left behind as it improves. While recent environment generation methods attempt to address this, they require domain-specific pipelines, rely on expensive or unreliable verifiers, and still produce static environments.
For AI agents to be useful beyond simple chat, they must hold sensitive user context such as calendars, credentials, health records, and financial data. We study whether the mere presence of such secrets in a model's context window introduces hidden correlations into the model's benign outputs, allowing reconstruction even when the model correctly refuses direct extraction.
Agentic reinforcement learning (RL) has become a critical stage in the post-training of large language models. Existing critic-free, group-relative methods estimate policy advantages from multiple rollouts, avoiding the substantial memory overhead of conventional proximal policy optimization (PPO) and achieving strong performance on long-horizon interactive tasks.
SWE‑bench Science is a benchmark for evaluating coding agents on scientific software engineering tasks, comprising 119 tasks from 98 GitHub repositories across 20 scientific domains. The tasks are grouped into Issue‑driven, Expert‑exploratory, and Engineering‑integration paradigms, and even the best agent, Claude Code with Opus‑5 (max), achieves a pass@1 below 50%. The study identifies four common failure mechanisms—lack of scientific knowledge, misguided exploration, incomplete repair coverage, and poor generalization—and shows that providing scientific guidance can both help and hinder repair depending on its alignment with the task.
The paper presents a method to align a language‑model‑based crowd agent with aggregate mobility data by fine‑tuning it to match observed destination compositions derived from origin‑to‑destination flows. The approach uses iterative proportional fitting to reweight the model’s destination distribution and corrects for dominant destination inflation by training a low‑rank adapter on resampled trajectories. Experiments on mobile network counts from two baseball games show a 25% reduction in destination‑share error while maintaining similar grid correlation across policies.
Multimodal LLMs can see a document, but they often can't read it reliably. Small text, tables, visual cues, and topological elements still trip them up under direct visual inference, even when the pag...