Tool use, function calling, orchestration and the protocols that let models act rather than only answer.
arXiv:2608.23030v1 Announce Type: new Abstract: We study how an AI system can identify and model other agents in its environment from observation alone, which is a capability necessary for cooperativ...
arXiv:2608.23152v1 Announce Type: new Abstract: Counterspeech effectively neutralizes the impact of online hate. Although prior work explores automated counterspeech generation, it largely emphasizes...
arXiv:2608.22960v1 Announce Type: new Abstract: Coding agents are increasingly evaluated not only by whether they solve a task, but also by how they execute it. However, existing process-level evalua...
arXiv:2608.22847v1 Announce Type: new Abstract: Vision-Language Models (VLMs) based GUI agents stand to benefit significantly from online reinforcement learning (RL). However, their training is bottl...
The paper proposes an information‑flow perspective on explainability, arguing that exposing reasons for observed effects is a positive flow of information that must be specified and verified. It introduces an epistemic temporal logic with counterfactual causes to formalize the requirement that agents gain knowledge about why an effect occurred, and presents an algorithm for checking finite‑state models against these specifications. A prototype implementation is evaluated on benchmarks, demonstrating the ability to distinguish explainable from unexplainable systems and to incorporate privacy constraints.
arXiv:2608.22026v1 Announce Type: new Abstract: Accurate simulation of the long-time evolution of systems governed by partial differential equations (PDEs) is central to scientific computing. Among e...
The paper introduces the Shared Recurrent Memory Transformer (SRMT), a decentralized multi‑agent reinforcement learning framework that uses a global memory workspace for agents to broadcast and query each other’s learned states. SRMT is evaluated on the Partially Observable Multi‑Agent Pathfinding (PO‑MAPF) problem, showing that shared memory enables emergent coordination even with minimal reward guidance and outperforms existing baselines on the Bottleneck task and scales competitively on larger POGEMA maps. The authors provide open‑source code for training and evaluation on GitHub.
arXiv:2608.23435v1 Announce Type: cross Abstract: Understanding a basketball game requires recognizing events, localizing actions, identifying players, and relating these to structured game knowledge...
InjecMEM introduces a memory injection attack that can steer the responses of large language model agents toward a desired output using only a single interaction, without needing read or edit access to the memory store. The attack leverages the retrieval‑then‑generate workflow of memory systems by crafting a retriever‑agnostic anchor with high‑recall topical cues and an adversarial command optimized through gradient‑based coordinate search. Experiments across various memory systems and backbone models show that InjecMEM reliably induces topic‑conditioned retrieval and targeted generation, remains effective even when memory drifts, and does not affect non‑target queries.
Meta-Ctrl is a constrained‑decoding framework that guarantees syntactic and semantic constraints in robot plans generated by large language models. It introduces meta‑tokens—a compact vocabulary of grounded actions—to enforce syntax at the token level and semantics (preconditions, goals, ordering) at the action level, dramatically reducing memory usage from over 107 TB to under 2 GB. The approach enables a small open‑weight LM to outperform GPT‑4 on subgoal success rates in the WAH‑NL LoTa‑Bench protocol and consistently improves performance on the Embodied Agent Interface, while real‑world tabletop robot experiments show every plan satisfies its preconditions and goals by construction.
arXiv:2602.03318v4 Announce Type: replace Abstract: Operations Research (OR) relies on expert-driven modeling--a slow and fragile process ill-suited to novel scenarios. While large language models (L...
arXiv:2608.22266v1 Announce Type: new Abstract: In the context of information seeking, conversational agents are undergoing an evolution from reactive tools to proactive, personalized assistants. A c...
CallScreenBench is a benchmark for evaluating small, on-device language models that act as phone secretaries, focusing on their ability to handle unknown inbound calls without a cooperative task. The benchmark measures owner endorsement through five call-and-note metrics, each paired with counter-metrics and uncertainty estimates, and includes guardedness diagnostics to identify safe, tool‑free proxies. Results across 4‑bit checkpoints of 0.6‑4 B parameter models show varying performance on service, recall, plausibility, and triage discrimination, highlighting trade‑offs between quality and guardedness.
arXiv:2510.03536v3 Announce Type: replace-cross Abstract: Multi-turn medical question answering (QA) aims to model realistic clinical diagnosis, where a doctor gathers patient information across mult...
ST‑EVO is a new framework for Large Language Model‑powered Multi‑Agent Systems that introduces a spatio‑temporal evolution of communication topologies. It uses a flow‑matching based scheduler to enable dialogue‑wise communication scheduling and incorporates uncertainty perception and self‑feedback to refine its scheduling over time. Experiments on nine benchmarks show that ST‑EVO outperforms existing methods, improving accuracy by roughly 5% to 25%.
arXiv:2608.22444v1 Announce Type: new Abstract: The unit of AI safety evaluation is still the individual model, yet language-model agents are increasingly deployed in interacting populations that rea...
DRAgent introduces a discriminative reasoning framework for Referring Expression Segmentation that replaces direct coordinate prediction with a two‑stage selection process. First, a detector generates candidate boxes; then a multimodal large language model (MLLM) discriminates the correct target through high‑recall screening and instance‑wise verification. The chosen box serves as a spatial prompt for a foundation segmentation model, and a self‑consistency‑filtered data pipeline is used for LoRA‑based fine‑tuning, yielding competitive results on RefCOCO, RefCOCO+, and RefCOCOg.
The study examines how multilingual large language models (LLMs) produce outputs that differ across sociocultural contexts, highlighting that identity labels and source-language cues can mislead assessments of cultural grounding. Using a human‑validated, multi‑agent audit on 89,253 outputs from 12 LLMs in English, French, and Chinese across 18 occupations and three task conditions, the authors find that bias representation varies systematically by language and task. Removing direct identity cues reduces identity‑label prediction in English and Chinese but not in French, and the source language’s cultural context consistently receives the highest relevance score, though this signal weakens after translation or name masking. "whyItMatters":"The findings show that surface cues can obscure true cross‑cultural patterns, underscoring the need for careful audit designs to avoid misleading conclusions about bias in multilingual LLMs."
arXiv:2608.23104v1 Announce Type: cross Abstract: Molecular science represents an important frontier for LLM-based agents. Unlike general agents that mainly operate over natural language, code, or we...
arXiv:2608.21412v1 Announce Type: new Abstract: Root cause analysis (RCA) in large datacenter networks is challenging because telemetry is noisy, partial, and asynchronous. Score-based approaches deg...