Tool use, function calling, orchestration and the protocols that let models act rather than only answer.
arXiv:2608.30451v1 Announce Type: new Abstract: Image-based 3D visual grounding is critical for embodied agents, yet existing benchmarks suffer from loose text-observation alignment and neglect tempo...
SimCRAFT is a model‑agnostic framework that distills remote sensing orchestration into a compact 7B‑scale model. It creates a large, constraint‑validated workflow planning corpus (SimRS‑14k) using a multi‑agent synthesis engine and a Mock Execution Engine, then fine‑tunes the model with Contextual Retrieval‑Augmented Fine‑Tuning (CRAFT) to reason analogically. Experiments show SimCRAFT‑7B outperforms open‑weight LLMs and rivals advanced closed‑source models, providing a lightweight, efficient baseline for autonomous remote sensing deployment.
arXiv:2608.29910v1 Announce Type: new Abstract: Interactive world models extend video generation from offline clip synthesis toward persistent simulation of interactive virtual worlds, enabling appli...
arXiv:2608.29800v1 Announce Type: new Abstract: Tool-using agents are being proposed for medical imaging, and their behaviour when a tool returns a false finding is largely unmeasured. We audit wheth...
arXiv:2603.27476v3 Announce Type: replace Abstract: AI-powered people search platforms are increasingly deployed for recruiting, sales prospecting, and professional networking, yet no standardized be...
SkillForge is a framework that breaks down formal code synthesis into reusable atomic skills, each handling a specific subtask such as specification inference, body synthesis, invariant generation, error diagnosis, or repair. A verification-driven harness coordinates these skills by submitting candidates to the Dafny verifier, diagnosing failures, and routing them deterministically to the appropriate repair skill until correctness is achieved or a budget is reached. On a curated benchmark, SkillForge outperforms state‑of‑the‑art agentic and iterative baselines, requiring fewer tokens and lower latency, with ablation studies showing each skill’s measurable contribution and rapid convergence.
SocialReasonBench is a new video‑multiple‑choice QA benchmark designed to test socially grounded reasoning in interactive narrative videos. It uses branching gameplay footage from *Detroit: Become Human*, where player choices create alternative social outcomes that can be verified against the game’s script and flowchart. The benchmark includes seven reasoning dimensions—such as intent recognition, emotional empathy, moral dilemma, counterfactual reasoning, and causal antecedent—and employs a multi‑agent pipeline to curate clips, ground answer labels, and generate theory‑guided questions with diagnostic distractors.
arXiv:2606.13120v2 Announce Type: replace Abstract: Search Agents -- large language models augmented with search tools -- have intensified the need for future-proof evaluation benchmarks. Existing be...
The paper introduces an ontology‑guided multi‑agent framework for extracting evaluation objects from academic review texts, addressing challenges such as abstractness, context‑dependency, and ambiguous type boundaries. The system combines candidate discovery, ontology‑constrained classification, and domain review, achieving high precision (90.33%) and recall (84.55%) and outperforming rule‑based and zero‑shot baselines. Ablation studies show that the multi‑agent workflow boosts recall and stability, while ontology‑based constraints improve fine‑grained classification and reduce category confusion.
arXiv:2605.30947v4 Announce Type: replace Abstract: LLM-based research agents have advanced rapidly in science and engineering, where research is organized around executable experiments, code, and qu...
Agent Zero Memory is a provenance‑aware long‑term memory system for large language model agents that distills user interactions into three parallel memory structures: an episodic timeline, an associative entity‑event knowledge graph, and a semantic, citation‑locked hierarchical documentary memory. Retrieval is performed via an intent gate, source router, and concurrent searches across the three systems, producing integrated, cited answers that exclude fabrication and require evidence the reader has opened. The system achieves state‑of‑the‑art performance on LongMemEval (95.60%) and LoCoMo (93.60%) while offering a favorable accuracy‑cost‑latency trade‑off across multiple backbone LLMs.
arXiv:2608.28824v1 Announce Type: new Abstract: Currently, in biological systems Neural Correlates of Consciousness (NCCs) are characterized in terms of EEG and FMRI signals. Unfortunately, this char...
The paper introduces MIMIC-DOS, a dataset derived from MIMIC-IV that focuses on ICU cases where patient symptoms and medical signs are discordant. It presents CARE, a privacy‑compliant multi‑stage agentic reasoning framework that uses a proprietary LLM to generate structured categories and transitions, while a local LLM performs evidence acquisition and decision‑making. In retrospective evaluations on MIMIC‑DOS, CARE outperforms other LLMs and agentic workflows, demonstrating stronger handling of conflicting clinical evidence while preserving patient privacy.
CAST is a critique‑aware training framework that transforms sparse task outcomes into action‑level supervision for both critique learning and policy optimization. By analyzing agent trajectories, CAST synthesizes structured rationales that explain action validity under partial observability, enabling the creation of richer training data. Fine‑tuned Qwen3‑family models trained with CAST show significant reliability gains, outperforming GPT‑OSS‑120B by over 10% on Retail tasks and improving Telehealth performance by 9% in an out‑of‑domain setting.
arXiv:2601.05107v2 Announce Type: replace Abstract: As LLM-based agents are increasingly used in long-term interactions, cumulative memory is critical for enabling personalization and maintaining sty...
arXiv:2601.21545v2 Announce Type: replace Abstract: Agentic systems accumulate persistent memory across sessions, tools, and tasks, and a later request must retrieve from it under two distinct constr...
arXiv:2608.30041v1 Announce Type: cross Abstract: Large language model agents place outputs from external skills into their execution context, allowing attacker-controlled data to influence later pri...
arXiv:2608.25920v2 Announce Type: replace Abstract: As large language model (LLM)-based multi-agent systems (MASs) are increasingly applied to long-horizon complex tasks, their reliability has emerge...
arXiv:2602.02475v2 Announce Type: replace Abstract: AI agents often fail in ways that are difficult to localize because executions are probabilistic, long-horizon, multi-agent, and mediated by noisy...
The paper introduces Uncertainty-Driven Replay Memory (UDRM), a new experience replay buffer for reinforcement learning that prioritizes storing transitions with high uncertainty estimates. Unlike traditional buffers that rely on temporal difference error or transition distributions, UDRM updates its contents based on uncertainty derived from the RL model during training. Experiments show that this uncertainty-aware buffer leads to higher rewards during training compared to other uncertainty-aware RL frameworks.