Tag
A 135-page survey paper organizing agentic reasoning for large language models along three layers—foundational, self-evolving, and collective multi-agent reasoning—spanning in-context and post-training optimization with real-world applications and open challenges.
DocSkill提出一种带技能编排策略的智能体框架,用于多语言多跳文档问答,通过MCTS离线探索轨迹并结合SFT/DPO训练策略,同时发布包含60K QA对、覆盖10种语言的MMHop-DocBench数据集。
The paper introduces MedCaseAgent, an agentic framework that turns unstructured clinical notes into publication-quality rare disease case reports by combining hierarchical patient data mining with iterative tool-integrated reasoning. It also releases the MedCase-150K corpus, MedCase-Bench evaluation suite, and an open-source model post-trained from Qwen3.5-9B that outperforms existing medical LLMs and agents.
LongHarness 是一个模型无关的智能体框架,通过「证据-结论协同演化」和「证据生命周期管理」机制,在超过 1 亿 token 的文档语料上进行有状态的证据管理,显著提升了长上下文推理任务的准确率(如 OfficeQA Pro V1 上从 15.8% 提升至 60.2%),并超越了 Claude Code 和 Codex 等现有智能体框架。
Meta Superintelligence Labs的研究者提出GitSwarm,一个通过共享Git仓库进行结构化持久记忆的异步多智能体系统,实现"compounding inference",使中间成果可被后续计算继承、扩展与组合;在IMOProofBench-Advanced上一次运行解出全部30题,在ProgramBench上取得79.4%的成绩。
Nous Research 发布了名为 MISSINGNO. 的编码与智能体推理模型,主打高 token 效率,并在 Nous Portal 上限时免费提供,可直接在 Hermes Agent 中使用。
Meta Superintelligence Labs introduces agentic meta-reasoning, an inference-time harness where a controller decides what work to run next in long agent runs, boosting ProgramBench from 63.7% to 71.5% with GPT-5.5 under the same workers and budget. Gains of 3.6-4.2 points hold across ProofBench, ARC-AGI-2 and LongCoT-mini when averaged over frontier models.
The paper proposes Wiki Foundation Model (WFM), a novel foundation model for scalable, agent-native representation and retrieval using LLM Wiki, which couples dense text with link structure, achieving strong performance on benchmarks and 10.5x training acceleration.
The paper proposes WFM, a Wiki Foundation Model for scalable, agent-native knowledge representation and retrieval, demonstrating improved performance in complex agentic reasoning tasks with a 10.5x training acceleration.
ARCHE is an autonomous agentic system that integrates reasoning and computational models to automate the discovery and validation of chemical reaction mechanisms, validated through various chemical scenarios.
This paper introduces an agentic framework combining LLMs and VLMs for consistent multi-instruction video editing across multiple shots, and proposes the MMLVE task and benchmark to evaluate performance.
DeAR is a decentralized agentic reasoning framework that enhances accuracy in knowledge-intensive reasoning tasks through capability grounding and collaborative thought navigation, outperforming centralized methods on multimodal benchmarks.
This paper presents a unified global-to-local paradigm for video anomaly detection, introducing a training-free framework (GtS) and a tool-augmented agentic reasoning method with reinforcement learning, along with a new benchmark VAGU-T and metric JeAUG.
QuantumMind presents an auditable agentic workflow that automatically generates and screens quantum speedup hypotheses using typed role-specialized actions and a deterministic validator.
Maxime Labonne highlights early fine-tunes of LFM2.5-2.6B, while Bad Theory Labs releases two open-weight models: BTL-4 35B, a frontier agentic reasoning model, and Macaw 2.7B, an on-device Mac agent, with notable BFCL v4 performance.
CastFSR is a Fast–Slow–Reflect agentic reasoning framework that leverages LLMs for context-aware time series forecasting, combining fast lightweight forecasters, slow deliberative reasoning, and reflective evaluation to improve forecasting accuracy and consistency.
Introduces agentic graph token reasoning, a method that recasts graph tokenization as part of an LLM's step-by-step reasoning process, allowing the model to dynamically choose graph views and granularity during inference. The method outperforms baselines across seven graph domains and transfers zero-shot to unseen domains.
This paper introduces LedgerMind, a provenance-constrained multimodal agentic reasoning framework that uses a Structured Evidence Ledger to ensure grounded, faithful reasoning in visual question answering, addressing failure patterns like hallucination and over-reasoning.
SCAIR introduces a training-free framework for reasoning over enterprise knowledge graphs by injecting schema-conditioned structural priors and enforcing schema-aware traversal during multi-hop reasoning, showing improvements on a real-world CMDB benchmark.
Legora introduces the Legora BAR benchmark for evaluating AI agentic reasoning in legal workflows, using real cases and the Legora harness to measure system-level performance.