Hyper-relational knowledge graphs (HKGs) extend traditional knowledge graphs by enriching triples with attribute–value qualifiers, enabling the fusion of multi-granular factual knowledge. The task of hyper-relational knowledge graph completion (HKGC) aims to infer plausible missing links by jointly exploiting information from both main triples and their qualifiers. Existing approaches mainly emphasize direct interactions between triples and qualifier pairs, but often overlook the monotonicity properties that naturally emerge from the fusion of coarse-grained and fine-grained knowledge. To address this, we propose the HyperMono model, which introduces a two-stage reasoning mechanism. In the first stage, triple-level reasoning produces coarse-grained inference signals, which are then refined in the second stage through triple+qualifier reasoning to achieve fine-grained predictions. This design leverages qualifiers to refine the candidate answer space derived from the main triple, providing a qualifier-aware prediction process that reduces answer drift in hyper-relational reasoning. To implement this property, HyperMono represents triples as cones and models each qualifier as a cone that incrementally shrinks the triple cone. Furthermore, HyperMono integrates neighborhood context information to semantically strengthen entity representations, thus realizing a richer fusion of structural and contextual knowledge. Extensive experiments on three real-world datasets under multiple evaluation scenarios demonstrate that HyperMono significantly outperforms state-of-the-art baselines. Datasets and code are available at the following website: https://github.com/zhiweihu1103/HKGC-HyperMono.

Large Language Models (LLMs) have become a core component of modern intelligent systems for knowledge-intensive applications. However, despite storing vast amounts of factual knowledge in their parameters, they often fail to effectively utilize this knowledge for bidirectional reasoning. A representative manifestation of this limitation is the “Reversal Curse”, where models trained on “A is B” cannot generalize to “B is A.” To systematically diagnose this deficiency, we design a knowledge graph-based structured recall evaluation framework that assesses whether LLMs can externalize internally stored knowledge into explicit relational triples. Evaluating six LLMs of varying scales and training strategies on a benchmark of 4000 entities with 84,142 Wikidata triples, we find that even the strongest models achieve less than 12% zero-shot structured recall accuracy on the model-specific benchmark, and recall rates remain critically low on a cross-validated intersection benchmark that controls for dataset variation. This finding suggests that the bottleneck lies not merely in knowledge storage, but in the model’s ability to actively retrieve and organize its parametric knowledge into structured representations that support inference. To address this, we propose the Structured Knowledge Recall (SKR) framework, a curriculum learning-inspired two-stage training paradigm that teaches models to first recall entity-centered knowledge graphs from their parameters and then perform reasoning over the explicitly recalled structures. Extensive experiments across four datasets and three model backbones demonstrate that SKR substantially mitigates the Reversal Curse, achieving over 92% reverse generalization accuracy while maintaining forward accuracy above 98%, and that endogenously recalled structured knowledge significantly outperforms externally injected knowledge, offering a new practical pathway for building more robust intelligent language systems.

Long-form question answering increasingly relies on retrieved evidence to make LLM outputs verifiable, with inline citations tracing claims to source documents. However, existing systems often attach citations that are topically related but insufficient to support their claims. We identify attribution ambiguity as a structural challenge: end-to-end generation must implicitly resolve combinatorial claim--document assignments, obscuring evidential boundaries and increasing the risk of evidence-boundary overrun, where claims exceed cited support. To address this challenge, we propose CAGE (Cognitive Attribution Graphs for Citation Generation), a two-stage framework that introduces an explicit cognitive attribution map before answer generation. CAGE first trains a plug-and-play Cognitive Map Induction Model to construct answer-centered support subgraphs, aligning each semantic answer unit with supporting documents through explicit relations. A Structured Citation Reasoning Model then realizes these units as sentence-level claims with map-aligned citations. Experiments on ASQA, ELI5, and ExpertQA show that CAGE achieves state-of-the-art performance, demonstrating the effectiveness of attribution-space contraction and map-guided citation generation.

World-model synthesis aims to turn interaction experience into an internal model of environment dynamics. Existing symbolic approaches often fit observed transitions or mixtures of local rules, but they do not produce a complete executable program that can run independently of the real environment. We present Mind-Studio, a framework that synthesizes executable pygame-style world models from state-action-next-state trajectories using large language models. Mind-Studio combines entropy-selected traces with a lightweight game skill file containing object, action, and static scene information extracted from screenshots. We evaluate synthesis quality with a K-step lookahead fidelity protocol that compares generated world-model rollouts against Real-ALE rollouts from the same state. On Montezuma's Revenge, Mind-Studio improves chosen-action next-state prediction from 0.3% for PoE-World to 48.7% while verifying 5 of 8 subgoals; across Alien, Assault, and Skiing, it achieves stronger branch-level fidelity than prior learned lookahead sources.

The rapid evolution of Large Language Models has catalyzed a surge in scientific idea production, yet this leap has not been accompanied by a matching advance in idea evaluation. The fundamental nature of scientific evaluation needs knowledgeable grounding, collective deliberation, and multi-criteria decision-making. However, existing idea evaluation methods often suffer from narrow knowledge horizons, flattened evaluation dimensions, and the inherent bias in LLM-as-a-Judge. To address these, we regard idea evaluation as a knowledge-grounded, multi-perspective reasoning problem and introduce , a deep innovation evaluation framework designed to emulate human-level idea assessment. We apply a heterogeneous deep knowledge search engine that retrieves and grounds dynamic evidence from diverse online sources. We further achieve review consensus with an innovation review board containing reviewers with distinct academic backgrounds, enabling a multi-dimensional decoupled evaluation across multiple metrics. We construct comprehensive datasets derived from authoritative peer-reviewed submissions to benchmark InnoEval. Experiments demonstrate that InnoEval can consistently outperform baselines in point-wise, pair-wise, and group-wise evaluation tasks, exhibiting judgment patterns and consensus highly aligned with human experts.

Chain-of-Thought (CoT) has unlocked advanced reasoning abilities of Large Language Models (LLMs) with intermediate steps, yet incurs prohibitive computational costs due to generation of extra tokens. Recent studies empirically show that compressing reasoning steps into latent states, or implicit CoT compression, offers a token-efficient alternative. However, the mechanism behind CoT compression remains unclear. In this paper, we provide the first theoretical analysis of the difficulty of learning to internalize intermediate reasoning steps. By introducing Order-r Interaction, we prove that the learning signal for high-order logical dependencies exponentially decays to solve irreducible problem, where skipping intermediate steps inevitably leads to high-order interaction barriers. To empirically validate this, we introduce NatBool-DAG, a challenging benchmark designed to enforce irreducible logical reasoning and eliminate semantic shortcuts. Guided by our theoretical findings, we propose ALiCoT (Aligned Implicit CoT), a novel framework that overcomes the signal decay by aligning latent token distributions with intermediate reasoning states. Experimental results demonstrate that ALiCoT successfully unlocks efficient reasoning: it achieves a 54.4x speedup while maintaining performance comparable to explicit CoT.

Existing benchmarks for deep research agents (DRAs) assess only single-shot outputs, ignoring a key question: can DRAs improve their reports when guided by feedback? To investigate this, we conduct a multi-turn evaluation of DRAs under two feedback settings: self-reflection, in which the agent revises its report without any external diagnostic signal, and process-level feedback, in which the agent receives guidance targeting gaps in its research strategy. To enable process-level feedback, we design Research Gap Inference (RGI), a method that analyzes patterns of satisfied and unsatisfied rubric criteria to infer research-process gaps. Our analysis reveals three key findings: (i) under self-reflection, agents incorporate and regress on rubric criteria at nearly equal rates, yielding negligible net improvement; (ii) a single round of process-level feedback yields substantial gains, raising the normalized score by approximately 8-15 points and yielding a roughly 35-40% incorporation rate; (iii) these gains do not compound over subsequent turns, as agents regress on up to 24% of previously satisfied criteria when rewriting the full report to address remaining gaps. Even with targeted guidance, reliable multi-turn improvement remains out of reach for the DRA architectures we evaluate. Our code and results are publicly available at https://github.com/sabharwalrishabh/Multi-Turn-Evaluation-of-DRAs.

Large language model (LLM)-based agents are increasingly used in interactive textual environments, from web navigation and code editing to tool use and long-horizon dialogue. Yet many remain largely reactive, mapping observations to actions without an explicit model of how these environments are structured and evolve. This motivates text world models (TWMs): transition models over textual states that, given a state and a candidate action, predict the resulting webpage, terminal output, API response, or user reply, thereby supporting planning, efficient learning, and principled evaluation. We systematically review text world models for LLM-based agents, organized around a formal framework and the agent lifecycle: (1) Foundations, defining text world models and characterizing them by state representation and grounding domain; (2) Construction, taxonomizing LLM-as-WM and code-as-WM paradigms and reviewing methods for building them; (3) Application, examining how world models support agents at training time through experience synthesis and at inference time through planning, verification, and adaptation; and (4) Evaluation, covering both evaluation of the world model itself and its use as an evaluation environment for agents. We aim to consolidate this rapidly developing area, clarify its design space, and highlight open challenges for future research.

Large language models (LLMs) risk retaining sensitive, copyrighted, or harmful information from their training data. Entity-level unlearning addresses this issue by removing all knowledge of a specific entity while preserving the model's overall capabilities. Existing approaches typically rely on full-model fine-tuning or prompt-based interventions, which can be computationally expensive or brittle when handling paraphrased queries. Recently, model editing has emerged as an efficient alternative for updating knowledge in LLMs, offering a promising direction for unlearning. However, existing editing techniques are typically designed for instance-level updates, modifying responses to specific attributes of an entity rather than eliminating all knowledge associated with the entity. In this paper, we investigate how editing techniques can be adapted for effective and efficient entity-level unlearning. To this end, we introduce a novel consistency-aware editing (CAE) framework. CAE aggregates a diverse set of prompts related to a target entity, including its attributes, relations, and adversarial paraphrases. It then jointly learns a low-rank update guided by a consistency regularizer that aligns the editing directions across prompts. This promotes robust and comprehensive forgetting while minimizing interference with unrelated knowledge. We further examine where different entities are stored within the model and how many diverse prompts are needed for successful unlearning. We evaluate CAE on two challenging benchmarks, RWKU and ToFU, and demonstrate that it (i) provides insights into how entity-level knowledge is internally represented and deleted in LLMs, (ii) significantly improves forgetting accuracy and robustness over traditional unlearning and editing baselines, and (iii) enables scalable entity removal using only tens of carefully selected prompts.

Large Language Models (LLMs) have demonstrated impressive capabilities across a wide range of NLP tasks, but they remain fundamentally stateless, constrained by limited context windows that hinder long-horizon reasoning. Recent efforts to address this limitation often augment LLMs with an external memory bank, yet most existing pipelines are static and heuristic-driven, lacking a learned mechanism for deciding what to store, update, or retrieve. We present Memory-R1, a reinforcement learning (RL) framework that equips LLMs with the ability to actively manage and utilize external memory through two specialized agents: a Memory Manager that learns structured operations, including ADD, UPDATE, DELETE, and NOOP; and an Answer Agent that pre-selects and reasons over relevant entries. Both agents are fine-tuned with outcome-driven RL (PPO and GRPO), enabling adaptive memory management with minimal supervision. With only 152 training QA pairs, Memory-R1 outperforms strong baselines and generalizes across diverse question types, three benchmarks (LoCoMo, MSC, LongMemEval), and multiple model scales (3B–14B).

Temporal reasoning over long, multi-session dialogues is a critical capability for conversational agents. As dialogue histories grow in length and accumulate noise, existing long-context models struggle to accurately identify temporally pertinent information, significantly impairing reasoning performance. To address this, we introduce **Memory-T1**, a framework that learns a time-aware memory selection policy using reinforcement learning (RL). It employs a coarse-to-fine strategy, first pruning the dialogue history into a candidate set with temporal and retriever filters, followed by an RL agent that selects the precise evidence. The RL training is guided by a multi-level reward function optimizing (i) accuracy, (ii) evidence grounding, and (iii) temporal consistency. This temporal consistency reward provides a dense signal by evaluating alignment at both the session-level (range proximity) and the utterance-level (evidence density), enabling the agent to resolve subtle chronological ambiguities. On the Time-Dialog benchmark, Memory-T1 boosts a 7B model to an overall score of 67.0\%, establishing a new state-of-the-art performance for open-source models and outperforming a 14B baseline by 10.2\%. Ablation studies show temporal consistency and evidence grounding rewards jointly contributing to a 15.0\% performance gain.Moreover, Memory-T1 maintains robustness up to 128k tokens, where baseline models collapse, proving effectiveness against noise in extensive dialogue histories.

Artificial intelligence is rapidly entering the core workflows of scientific research. Yet reliable scientific reasoning requires access to accumulated scientific knowledge with sufficient breadth, depth, and standardization. Current AI scientists typically assemble scientific knowledge through workflow- and discipline-specific pipelines, which provide incomplete coverage, leave relations implicit, and make knowledge acquisition pathways fragmented. Here we present SciAtlas, a shared, machine-actionable cross-disciplinary scholarly knowledge infrastructure that integrates evidential, conceptual, disciplinary, expertise, and normative layers under a shared schema. SciAtlas further achieves a unified neuro-symbolic retrieval mechanism that grounds heterogeneous research objects, propagates relevance across the scholarly topology, and projects the resulting relevance field into the context required by each scientific workflow. Across three representative workflows, SciAtlas broadens trajectory reconstruction by recovering overlooked research branches, deepens opportunity discovery by uncovering underexplored bottlenecks and cross-domain insights, and strengthens innovation assessment by integrating evidence, expertise, and evaluation signals. Across three representative workflows, SciAtlas broadens trajectory reconstruction by recovering overlooked stages and branches, deepens opportunity discovery by uncovering underexplored bottlenecks and cross-domain connections, and standardizes innovation assessment by integrating evidence, expertise and evaluation signals. Extensive evaluations validate the foundational capabilities underpinning it as reusable knowledge infrastructure for knowledge-intensive scientific research.

A large language model (LLM) agent is a model coupled to a harness: parametric weights that hold knowledge and reasoning ability, wrapped in an external apparatus of memory, tools, and orchestration through which the model perceives and acts. Progress over the past years has come from pushing on both, with stronger models and richer harnesses, but the design decision that quietly governs an agent’s behavior is where competence should live: which knowledge, skills, and control to internalize into the model’s parameters, and which to externalize onto the surrounding harness. These decisions increasingly determine whether an agent is reliable, efficient, and able to keep improving. We investigate this core problem through the lens of the Theory of Agent (ToA), which views an LLM agent not as a task executor but as an epistemic decision-maker in terms of reasoning or acting at every step. The two means it has for that epistemic uncertainty reduction, namely reasoning over what it already holds and acting on the world to obtain what it does not, are the run-time faces of internalization and externalization: reasoning draws on competence that has been absorbed into the model, while acting exploits competence that remains in the harness. We review this internalization and externalization across five chapters. Foundations sets up the basic picture: an agent that, at each step, decides whether to answer from what it already knows or reach outside for what it does not, and what separates a good version of that decision from a wasteful one. Learning examines how training places competence on either side of the boundary, spanning internalization of memory, reasoning, tool-use policy, and world models, and externalization of memory, skills, and orchestration. Aligning treats calibration as the connecting thread across an agent’s cognitive boundaries, efficiency constraints, human preferences, environment constraints, and other agents. Evolving studies how agents revise both sides of the boundary over time, from model-side and harness-side updates to agent–environment co-evolution. Evaluation reinterprets metrics and benchmarks by the epistemic effort they expose and argues for measuring the boundary decision rather than task completion alone. We conclude with the open problems that follow once internalization and externalization are treated as the governing axis of agent design.

The rapid growth of open and structured RDF data on the Web has promoted the development of dataset search as an important research topic. The core function of existing systems is ad hoc dataset retrieval (AHDR) based on the metadata of datasets, which contains limited information and often suffers from quality issues. To overcome the limitations, in this article, we systematically investigate content-based AHDR to exploit the actual RDF data in datasets. We address three main tasks of content-based AHDR with novel methods for handling the large size and complex structure of RDF data to facilitate dataset retrieval, deduplication, and snippet extraction. These methods are integrated into an online and open-source prototype called Caddie. The effectiveness and practicability of its components are evaluated on a public test collection and by a user study.

Recent studies have explored the capabilities of large language models (LLMs) in solving knowledge-intensive mathematical reasoning problems. However, existing benchmarks predominantly involve static theorems that LLMs have encountered during pretraining, failing to assess dynamic knowledge integration. In this work, we introduce TaxReasoning, a novel benchmark designed to evaluate LLMs’ abilities in real-world tax calculation scenarios. These tasks require not only mathematical reasoning and numerical computation, but also the extraction and application of complex, frequently updated tax regulations. Through extensive experiments with state-of-the-art LLMs using diverse prompting strategies and knowledge augmentation techniques, we uncover substantial limitations in their ability to handle dynamic, knowledge-intensive questions—primarily due to missing domain-specific knowledge and ineffective retrieval. Even the best-performing models fall significantly short of human-level performance. Our analysis points to key avenues for improvement, including enhancing LLMs' reasoning capabilities, developing more effective knowledge summarization techniques, and improving retrieval strategies. TaxReasoning offers a critical testbed for advancing LLMs in dynamic knowledge-intensive domains.

Terminal agents extend Large Language Models with the ability to execute tasks directly in command-line environments, but their progress is bottlenecked by the scarcity of high-quality training data. Existing approaches bootstrap from partial sources such as human-defined seeds or GitHub repositories to instantiate one component and then complete the rest, producing tasks confined to narrow seed distributions, environments misaligned with task semantics, and inefficient trajectories from unguided exploration. To address these limitations, we introduce Terminal-World, a fully automated pipeline that uses agent skills as the central synthesis primitive, which jointly encode what to accomplish, when to apply (preconditions and environment state), and how to execute, enabling task instructions, environments, and teacher trajectories to be co-derived. To further broaden the synthesis space, Terminal-World composes skills into skill teams and skill graphs for multi-role and cross-domain task synthesis. Using this pipeline, we construct 5,723 training environments and train Terminal-World-8B/14B/32B, evaluated across 6 benchmarks where the Terminal-World series consistently outperforms terminal-agent baselines. Notably, using the same teacher model and only 1.2

The identification and filtration of malicious texts in social media environments represent a significant technical challenge aimed at protecting users from online violence and disinformation. This complexity stems from the diversity and innovativeness of social media texts, which include unique expressions and special sentence structures. Particularly, malicious texts in interrogative forms pose alignment challenges with traditional corpora due to existing methods' failure to exploit the text's deep global semantic representations. This issue is compounded by the scant research on Chinese texts, leading to inefficiencies in recognition accuracy. To mitigate these challenges, we introduce an innovative framework based on a Global Contrastive Semantic Network (GCSN), designed to enhance malicious text recognition efficiency and accuracy by deeply learning global semantic knowledge. It comprises an encoder for global semantic information modelling and a graph-matching network for semantic similarity evaluation between question pairs, enabling the accurate identification and filtering of malicious texts with complex structures. Furthermore, we introduce a semantic consistency-based data augmentation method (COMBINE), using real-world data to generate balanced positive and negative samples, enriching the dataset and enhancing the model's ability to distinguish semantic consistency through contrastive learning. Experimental validation on two Chinese datasets demonstrates our model's exceptional performance, affirming its applicationa value in social media malicious text recognition. Our code is available at https://github.com/Wxy13131313131/GCSN-COMBINE

Recent research has explored the integration of knowledge graphs (KGs) with large language models (LLMs) to enhance their performance on downstream knowledge-intensive tasks, particularly knowledge graph question answering (KGQA). Existing approaches primarily combine LLMs with KGs through retrieval-augmented generation (RAG)-based, agent-based, and SPARQL-based methods. Although these methods have achieved notable success, they still suffer from several limitations, including structural information loss, unfaithful reasoning, and limited flexibility and generalization. To address these challenges, this paper proposes KG2Code, a novel approach that transforms knowledge graphs into a code-based representation, preserving structural semantics while naturally aligning with the code-aware pretraining of modern LLMs. Based on KG2Code, KG2Code-QA is further introduced as a KGQA framework that formulates KGQA as a code generation task. This formulation enables the generation of verifiable reasoning traces and executable code, thereby substantially mitigating the impact of hallucinations. In addition, an automated pipeline is developed to construct a large-scale, high-quality code corpus for effectively training open-source LLMs on KG2Code-QA. After training, LLMs are able to perform KGQA in zero-shot scenarios. Extensive experiments demonstrate that the proposed approach significantly outperforms existing KG-enhanced LLM methods for KGQA, while exhibiting strong generalization to unseen KGs. The code and data are available at Github.