Fact-checking has transitioned from traditional semantic matching toward active reasoning powered by Large Language Models (LLMs), manifesting as the interplay between internal parametric knowledge and external grounded evidence. However, this paradigm introduces two critical challenges: (i) excessive reliance on external evidence even when it conflicts with correct internal judgment, and (ii) uncritical acceptance of a consensus even when both sources are erroneous. To address these limitations, we propose a Multi-Agent Arbitration Framework where a meta-cognitive Adjudication Agent (AA) orchestrates a Knowledge Agent (KA) and an Evidence Agent (EA). Through hybrid reinforcement and imitation learning, the AA perceives conflict signals between KA and EA to adaptively calibrate judgmental autonomy against evidentiary support. Upon detecting a potentially erroneous consensus, the AA invokes a Reflective Rejection mechanism to scrutinize and disrupt deceptive alignment. Empirical evaluations on Snopes and PolitiFact demonstrate that our framework consistently outperforms state-of-the-art baselines. Notably, it effectively mitigates over-reliance errors while maintaining high fidelity in correct-consensus scenarios, recovering 26.2% of dual-error cases where static systems typically succumb to superficial consistency. Our code is publicly available at https://github.com/Zhang-Gorden/Perceive-Then-Adjudicate.

We study the problem of personalization in large language models (LLMs). Prior work predominantly represents user preferences as implicit, model-specific vectors or parameters, yielding opaque “black-box” profiles that are difficult to interpret and transfer across models and tasks. In contrast, we advocate natural language as a universal, model- and task-agnostic interface for preference representation. The formulation leads to interpretable and reusable preference descriptions, while naturally supporting continual evolution as new interactions are observed. To learn such representations, we introduce a two-stage training framework that combines supervised fine-tuning on high-quality synthesized data with reinforcement learning to optimize long-term utility and cross-task transferability. Based on this framework, we develop AlignXplore+, a universal preference reasoning model that generates textual preference summaries. Experiments on nine benchmarks show that our 8B model achieves state-of-the-art performance—outperforming substantially larger open-source models—while exhibiting strong transferability across tasks, model families, and interaction formats. Code and dataset are accessible at https://github.com/AntResearchNLP/AlignX-Family/tree/main/AlignXplorePlus.

Answer-only reinforcement learning (RL) trains reasoning models to solve fully specified problems, but many realistic queries omit a premise needed for a unique answer. In this setting, the useful response is not always refusal: the model should ask for the missing premise, condition its answer on the unknown quantity, or abstain when no informative conditional response is available. We present Ask-Condition-Abstain Reinforcement Learning (ACA-RL), a data-augmented RL framework for this setting. Its reasoning-graph-guided pipeline converts well-posed problems into missing-premise training instances with localized gap annotations; ACA-RL then trains on these instances with a structured reward over five observable response behaviors. We also introduce the Missing-Premise Benchmark (MPB), a 274-instance human-verified benchmark spanning mathematical, logical, and real-world word problems. Across Qwen3 and Llama models, ACA-RL consistently improves on MPB while preserving competitive performance on well-posed reasoning tasks. This work supports a new mission for NLP evaluation: measuring whether models can recognize when a task is underdetermined and handle uncertainty, not only whether they can answer fully specified questions.

Accurate comprehension and controllable generation of emotion and rhetoric are pivotal for enhancing the reasoning capabilities of large language models (LLMs). Existing studies mostly rely on external optimizations, lacking in-depth exploration of internal representation mechanisms, thus failing to achieve fine-grained steering at the neuron level. A handful of works on neurons are confined to emotions, neglecting rhetoric neurons and their intrinsic connections. Traditional neuron masking also exhibits counterintuitive phenomena, making reliable verification of neuron functionality infeasible. To address these issues, we systematically investigate the neurons representation mechanisms and inherent associations of 6 emotion categories and 4 core rhetorical devices. We propose a neuron identification framework that integrates multi-dimensional screening, and design an adaptive masking method incorporating dynamic filtering, attenuation masking, and feedback optimization, enabling reliable functional validation of neuron functionality. Through neuron regulation, we achieve directed induction of non-target sentences and enhancement of emotion tasks via rhetoric neurons. Experiments on 5 commonly used datasets validate the effectiveness of our method, providing a novel paradigm for the fine-grained steering of emotion and rhetoric expressions in LLMs.

Traditional industrial agents rely on modular pipelines, including Router, Retriever, Planner, Executor, Responder, Reviewer and so on, which inevitably fracture into a labyrinth of ad-hoc patches, leading to cascading errors and high latency. We propose OneModel, an applicable paradigm shift from external workflows to internalized knowledge representation. Unlike modular systems that slice fluid user intents into static steps, OneModel consolidates complex business logic and SOPs directly into the model’s parameters.Through Continual Pre-training (CPT) and logic-compilation SFT, we transform fragmented business rules into the model’s intuitive reasoning within a unified attention space. Deployed in our global financial service system, OneModel effectively breaks the impossible triangle of latency, accuracy, and complexity. Online A/B testing demonstrates end-to-end latency reduction of more than 50% (18.7s → 8s) while the Intelligent Resolution Rate (IRR) jumps from 64.3% to 83.3%. The results demonstrate our paradigm OneModel effectively replaces brittle engineering logic with internalized cognitive intuition, offering a scalable and future-proof blueprint for transitioning industrial agents from complex, error-prone workflow to unified model architectures.