Advanced large language models (LLMs) frequently reflect in reasoning chain-of-thoughts (CoTs), where they self-verify the correctness of current solutions and explore alternatives. However, given recent findings that LLMs detect limited errors in CoTs, how reflection contributes to empirical improvements remains unclear. To analyze this issue, in this paper, we present a minimalistic reasoning framework to support basic self-verifying reflection for small transformers without natural language, which ensures analytic clarity and reduces the cost of comprehensive experiments. Theoretically, we prove that self-verifying reflection guarantees improvements if verification errors are properly bounded. Experimentally, we show that tiny transformers, with only a few million parameters, benefit from self-verification in both training and reflective execution, reaching remarkable LLM-level performance in integer multiplication and Sudoku. Similar to LLM results, we find that reinforcement learning (RL) improves in-distribution performance and incentivizes frequent reflection for tiny transformers, yet RL mainly optimizes shallow statistical patterns without faithfully reducing verification errors. In conclusion, integrating generative transformers with discriminative verification inherently facilitates CoT reasoning, regardless of scaling and natural language.

Large Language Models (LLMs) hallucinate, generating factually incorrect yet confident assertions. We argue this stems from the Transformer's Softmax function, which creates "Artificial Certainty" by collapsing ambiguous attention scores into a single probability distribution, discarding uncertainty information at each layer. To fix this, we introduce the Credal Transformer, which replaces standard attention with a Credal Attention Mechanism (CAM) based on evidential theory. CAM produces a "credal set" (a set of distributions) instead of a single attention vector, with the set's size directly measuring model uncertainty. We implement this by re-conceptualizing attention scores as evidence masses for a Dirichlet distribution: sufficient evidence recovers standard attention, while insufficient evidence yields a diffuse distribution, representing ambiguity. Empirically, the Credal Transformer identifies out-of-distribution inputs, quantifies ambiguity, and significantly reduces confident errors on unanswerable questions by abstaining. Our contribution is a new architecture to mitigate hallucinations and a design paradigm that integrates uncertainty quantification directly into the model, providing a foundation for more reliable AI.

With the widespread use of multi-modal Large Language models (MLLMs), safety issues have become a growing concern. Multi-turn dialogues, which are more common in everyday interactions, pose a greater risk than single prompts; however, existing benchmarks do not adequately consider this situation. To encourage the community to focus on the safety issues of these models in multi-turn dialogues, we introduce SafeMT, a benchmark that features dialogues of varying lengths generated from harmful queries accompanied by images. This benchmark consists of 10,000 samples in total, encompassing 17 different scenarios and four jailbreak methods. Additionally, we propose Safety Index (SI) to evaluate the general safety of MLLMs during conversations. We assess the safety of 17 models using this benchmark and discover that the risk of successful attacks on these models increases as the number of turns in harmful dialogues rises. This observation indicates that the safety mechanisms of these models are inadequate for recognizing the hazard in dialogue interactions. We propose a dialogue safety moderator capable of detecting malicious intent concealed within conversations and providing MLLMs with relevant safety policies. Experimental results from several open-source models indicate that this moderator is more effective in reducing multi-turn ASR compared to existed guard models.

Precise attribute intensity control--generating Large Language Model (LLM) outputs with specific, user-defined attribute intensities--is crucial for AI systems adaptable to diverse user expectations. Current LLM alignment methods, however, typically provide only directional or open-ended guidance, failing to reliably achieve exact attribute intensities. We address this limitation with three key designs: (1) reformulating precise attribute intensity control as a target-reaching problem, rather than simple maximization; (2) training a lightweight value function via temporal-difference learning to predict final attribute intensity scores from partial generations, thereby steering LLM outputs; and (3) employing gradient-based interventions on hidden representations to navigate the model precisely towards specific attribute intensity targets. Our method enables fine-grained, continuous control over attribute intensities, moving beyond simple directional alignment. Experiments on LLaMA-3.2-3b and Phi-4-mini confirm our method's ability to steer text generation to user-specified attribute intensities with high accuracy. Finally, we demonstrate efficiency enhancements across three downstream tasks: preference data synthesis, Pareto frontier approximation and optimization, and distillation of aligned behaviors for intervention-free inference. Our code is available on https://github.com/Pre-Control/pre-control

Chatbot providers (e.g., OpenAI) rely on tiered subscription schemes to generate revenue, offering basic models for free users, and advanced models for paying subscribers. However, a finer-grained pay-to-unlock scheme for premium features (e.g., math, coding) is thought to be more economically viable for the providers. Such a scheme requires a feature-locking technique (FLoTE) which is (i) effective in refusing locked features, (ii) utility-preserving for unlocked features, (iii) robust against evasion or unauthorized credential sharing, and (iv) scalable to multiple features and users. However, existing FLoTEs (e.g., password-locked models) are not robust or scalable. We present Locket, the first robust and scalable FLoTE to enable pay-to-unlock schemes. Locket uses a novel merging approach to attach adapters to an LLM for refusing unauthorized features. Our comprehensive evaluation shows that Locket is effective ($100$% refusal on locked features), utility-preserving ($\leq 7$% utility degradation in unlocked features), robust ($\leq 5$% attack success rate), and scales to multiple features and clients.

Multilingual domain adaptation (ML-DA) is widely used to learn new domain knowledge across languages into large language models (LLMs). Although many methods have been proposed to improve domain adaptation, the mechanisms of multilingual knowledge acquisition, how domain knowledge is learned within a language and transferred across languages, remain underexplored. This gap leads to suboptimal performance, particularly in low-resource settings. This work examines the learning dynamics of LLMs during ML-DA. Because prior ML-DA studies often train and evaluate on datasets with mismatched knowledge coverage, we propose AdaXEval, an adaptive evaluation method that builds multiple-choice QA datasets from the same bilingual domain corpus used for training, thereby directly studying multilingual knowledge acquisition. Through continual training of LLMs with diverse data recipes, we track how LLMs acquire domain facts and pinpoint the mechanism behind the transformation process from domain training data to knowledge. Our experiments on a 13B English-Japanese bilingual LLM reveal that cross-lingual transfer remains challenging despite a high-quality bilingual corpus. The code has been released.

The evaluation of LLMs' creativity represents a crucial research domain, though challenges such as data contamination and costly human assessments often impede progress. Drawing inspiration from human creativity assessment, we propose PACE, asking LLMs to generate Parallel Association Chains to Evaluate their creativity. PACE minimizes the risk of data contamination and offers a straightforward, highly efficient evaluation, as evidenced by its strong correlation with Chatbot Arena Creative Writing rankings (Spearman's $ρ= 0.739$, $p < 0.001$) across various proprietary and open-source models. A comparative analysis of associative creativity between LLMs and humans reveals that while high-performing LLMs achieve scores comparable to average human performance, professional humans consistently outperform LLMs. Furthermore, linguistic analysis reveals that both humans and LLMs exhibit a trend of decreasing concreteness in their associations, and humans demonstrating a greater diversity of associative patterns.

Text-attributed graphs are widely used across domains, offering rich opportunities for zero-shot learning via graph-text alignment. However, existing methods struggle with tasks requiring fine-grained pattern recognition, particularly on heterophilic graphs. Through empirical and theoretical analysis, we identify an \textbf{over-abstraction problem}: current approaches operate at excessively large hyperbolic radii, compressing multi-scale structural information into uniform high-level abstractions. This abstraction-induced information loss obscures critical local patterns essential for accurate predictions. By analyzing embeddings in hyperbolic space, we demonstrate that optimal graph learning requires \textbf{faithful preservation} of fine-grained structural details, better retained by representations positioned closer to the origin. To address this, we propose \textbf{H4G}, a framework that systematically reduces embedding radii using learnable block-diagonal scaling matrices and Möbius matrix multiplication. This approach restores access to fine-grained patterns while maintaining global receptive ability with minimal computational overhead. Experiments show H4G achieves state-of-the-art zero-shot performance with \textbf{12.8\%} improvement on heterophilic graphs and \textbf{8.4\%} on homophilic graphs, confirming that radius reduction enables faithful multi-scale representation for advancing zero-shot graph learning.

We introduce ToPolyAgent, a multi-agent AI framework for performing coarse-grained molecular dynamics (MD) simulations of topological polymers through natural language instructions. By integrating large language models (LLMs) with domain-specific computational tools, ToPolyAgent supports both interactive and autonomous simulation workflows across diverse polymer architectures, including linear, ring, brush, and star polymers, as well as dendrimers. The system consists of four LLM-powered agents: a Config Agent for generating initial polymer-solvent configurations, a Simulation Agent for executing LAMMPS-based MD simulations and conformational analyses, a Report Agent for compiling markdown reports, and a Workflow Agent for streamlined autonomous operations. Interactive mode incorporates user feedback loops for iterative refinements, while autonomous mode enables end-to-end task execution from detailed prompts. We demonstrate ToPolyAgent's versatility through case studies involving diverse polymer architectures under varying solvent condition, thermostats, and simulation lengths. Furthermore, we highlight its potential as a research assistant by directing it to investigate the effect of interaction parameters on the linear polymer conformation, and the influence of grafting density on the persistence length of the brush polymer. By coupling natural language interfaces with rigorous simulation tools, ToPolyAgent lowers barriers to complex computational workflows and advances AI-driven materials discovery in polymer science. It lays the foundation for autonomous and extensible multi-agent scientific research ecosystems.

Large language models often mishandle psychiatric emergencies, offering harmful or inappropriate advice and enabling destructive behaviors. This study evaluated the Verily behavioral health safety filter (VBHSF) on two datasets: the Verily Mental Health Crisis Dataset containing 1,800 simulated messages and the NVIDIA Aegis AI Content Safety Dataset subsetted to 794 mental health-related messages. The two datasets were clinician-labelled and we evaluated performance using the clinician labels. Additionally, we carried out comparative performance analyses against two open source, content moderation guardrails: OpenAI Omni Moderation Latest and NVIDIA NeMo Guardrails. The VBHSF demonstrated, well-balanced performance on the Verily Mental Health Crisis Dataset v1.0, achieving high sensitivity (0.990) and specificity (0.992) in detecting any mental health crises. It achieved an F1-score of 0.939, sensitivity ranged from 0.917-0.992, and specificity was >= 0.978 in identifying specific crisis categories. When evaluated against the NVIDIA Aegis AI Content Safety Dataset 2.0, VBHSF performance remained highly sensitive (0.982) and accuracy (0.921) with reduced specificity (0.859). When compared with the NVIDIA NeMo and OpenAI Omni Moderation Latest guardrails, the VBHSF demonstrated superior performance metrics across both datasets, achieving significantly higher sensitivity in all cases (all p < 0.001) and higher specificity relative to NVIDIA NeMo (p < 0.001), but not to OpenAI Omni Moderation Latest (p = 0.094). NVIDIA NeMo and OpenAI Omni Moderation Latest exhibited inconsistent performance across specific crisis types, with sensitivity for some categories falling below 0.10. Overall, the VBHSF demonstrated robust, generalizable performance that prioritizes sensitivity to minimize missed crises, a crucial feature for healthcare applications.