Molecule generation is key to drug discovery and materials science, enabling the design of novel compounds with specific properties. Large language models (LLMs) can learn to perform a wide range of tasks from just a few examples. However, generating valid molecules using representations like SMILES is challenging for LLMs in few-shot settings. In this work, we explore how LLMs can generate 100% valid molecules. We evaluate whether LLMs can use SELFIES, a representation where every string corresponds to a valid molecule, for valid molecule generation but find that LLMs perform worse with SELFIES than with SMILES. We then examine LLMs' ability to correct invalid SMILES and find their capacity limited. Finally, we introduce SmiSelf, a cross-chemical language framework for invalid SMILES correction. SmiSelf converts invalid SMILES to SELFIES using grammatical rules, leveraging SELFIES' mechanisms to correct the invalid SMILES. Experiments show that SmiSelf ensures 100% validity while preserving molecular characteristics and maintaining or even enhancing performance on other metrics. SmiSelf helps expand LLMs' practical applications in biomedicine and is compatible with all SMILES-based generative models. Code is available at https://github.com/wentao228/SmiSelf.

Large language models (LLMs) trained with reinforcement objectives often achieve superficially correct answers via shortcut strategies, pairing correct outputs with spurious or unfaithful reasoning and degrading under small causal perturbations. We introduce Causally-Enhanced Policy Optimization (CE-PO), a drop-in reward-shaping framework that augments policy optimization with a differentiable proxy for causal coherence along the generation pathway from prompt (Z) to rationale (X) to answer (Y). CE-PO estimates model-internal influence with Jacobian-based sensitivities, counterfactually hardens these signals to suppress nuisance cues, and fuses the resulting coherence score with task-accuracy feedback via a Minkowski (power-mean) combiner, exposing a single tunable between accuracy and coherence trade-off. The unified reward integrates with PPO/GRPO without architectural changes. Across reasoning benchmarks and causal stress tests, CE-PO reduces reward hacking and unfaithful chain-of-thought while improving robustness to correlation-causation flips and light counterfactual edits, all at near-parity accuracy. Experimental results across 4 datasets show that CE-PO improves accuracy over baselines by 5.49% on average (up to 9.58%), while improving robustness to correlation-causation flips and light counterfactual edits.

Understanding uncertainty in large language models remains a fundamental challenge, particularly in creative tasks where multiple valid outputs exist. We present a geometric framework using credal sets - convex hulls of probability distributions - to quantify and decompose uncertainty in neural text generation, calibrated against human creative variation. Analyzing 500 creative writing prompts from the WritingPrompts dataset with 10 unique human continuations each, we evaluate four language models across five decoding strategies, generating 100,000 stories. Our credal set analysis reveals substantial gaps in capturing human creative variation, with the best model-human calibration reaching only 0.434 (Gemma-2B with temperature 0.7). We decompose total uncertainty into epistemic and aleatoric components, finding that the choice of decoding strategy contributes 39.4% to 72.0% of total epistemic uncertainty. Model scale shows weak correlation with calibration quality and no significant difference exists between base and instruction-tuned models in calibration quality. Our geometric framework provides actionable insights for improving generation systems for human-AI creative alignment. We release our complete experimental framework.

The rise of deep learning has greatly advanced human behavior monitoring using wearable sensors, particularly human activity recognition (HAR). While deep models have been widely studied, most assume stationary data distributions - an assumption often violated in real-world scenarios. For example, sensor data from one subject may differ significantly from another, leading to distribution shifts. In continual learning, this shift is framed as a sequence of tasks, each corresponding to a new subject. Such settings suffer from catastrophic forgetting, where prior knowledge deteriorates as new tasks are learned. This challenge is compounded by the scarcity and inconsistency of labeled data in human studies. To address these issues, we propose CLAD-Net (Continual Learning with Attention and Distillation), a framework enabling wearable-sensor models to be updated continuously without sacrificing performance on past tasks. CLAD-Net integrates a self-supervised transformer, acting as long-term memory, with a supervised Convolutional Neural Network (CNN) trained via knowledge distillation for activity classification. The transformer captures global activity patterns through cross-attention across body-mounted sensors, learning generalizable representations without labels. Meanwhile, the CNN leverages knowledge distillation to retain prior knowledge during subject-wise fine-tuning. On PAMAP2, CLAD-Net achieves 91.36 percent final accuracy with only 8.78 percent forgetting, surpassing memory-based and regularization-based baselines such as Experience Replay and Elastic Weight Consolidation. In semi-supervised settings with only 10-20 percent labeled data, CLAD-Net still delivers strong performance, demonstrating robustness to label scarcity. Ablation studies further validate each module's contribution.

Retrieval-augmented generation agents development is hindered by the lack of process-level supervision to effectively guide agentic capabilities like task decomposition, retriever invocation, and stepwise decision-making. While reinforcement learning offers a potential solution, it suffers from sparse rewards and the limited reasoning capabilities of large language models (LLMs). Meanwhile, existing data synthesis methods only produce chain-of-thought rationales and fail to model environmental interactions. In this paper, we propose EviPath, an evidence-anchored reasoning path synthesis paradigm for RAG agent development. EviPath comprises: (i) Abductive Subtask Planning, which decomposes the problem into sub-questions and iteratively plans an optimal solution path based on the dependencies between them; (ii) Faithful Sub-question Answering, which uses supporting evidence to construct a proxy environment to generate reasoning thoughts and answers for each sub-question; and (iii) Conversational Fine-Tuning, which formats the complete agent-environment interaction trajectory into a dialogue format suitable for Supervised Fine-Tuning. EviPath allows LLMs to learn complex reasoning and tool-use capabilities directly from synthesized data. Extensive experiments on widely-used question-answering benchmarks show that an 8B parameter model trained with EviPath-synthesized data significantly and consistently outperforms state-of-the-art baselines with a double-digit absolute EM gain of 14.7% in open-domain question answering. Retrieval-augmented generation (RAG) agents, powered by large language models (LLMs) (Guo et al., 2025), can autonomously gather external knowledge and answer complex, multi-hop questions. Compared to vanilla RAG systems (Lewis et al., 2020), RAG agents minimize the need for human intervention, and adapt readily to downstream applications like math problem solving (Zhu et al., 2025), code generation (Zhang et al., 2023), and financial analysis (Wang et al., 2025c). Despite their promise, RAG agents are hard to develop since ground truth reasoning trajectories are unavailable. Mainstream multi-hop question answering datasets Y ang et al. (2018); Ho et al. (2020); Trivedi et al. (2022) provide final answers and supporting facts, while lacking step-wise supervision that is crucial to equip LLMs with agentic behaviors like question decomposition, search query reformulation, and plan refinement.

We introduce DafnyCOMP, a benchmark for evaluating large language models (LLMs) on compositional specification generation in Dafny. Unlike prior benchmarks that focus on single-function tasks, DafnyCOMP targets programs composed of multiple interacting functions with data dependencies, requiring reasoning across component boundaries. The benchmark consists of 300 automatically synthesized multi-function programs. We evaluate several state-of-the-art LLM families and find that, while they perform well on single-function verification, their performance drops sharply on compositional tasks. Analysis reveals systematic failures in cross-functional reasoning, including fragile specifications, misalignment between implementations and proofs, and unstable reasoning. DafnyCOMP thus provides a diagnostic tool for measuring progress toward reliable, verifiable, and compositional code generation with LLMs.

Abstract--End-of-Life (EoL) phones significantly exacerbate global e-waste challenges due to their high production volumes and short lifecycles. Disassembly is among the most critical processes in EoL phone recycling. However, it relies heavily on human labor due to product variability. Consequently, the manual process is both labor-intensive and time-consuming. In this paper, we propose a low-cost, easily deployable automated and selective disassembly and sorting system for EoL phones, consisting of three subsystems: an adaptive cutting system, a vision-based robotic sorting system, and a battery removal system. The system can process over 120 phones per hour with an average disassembly success rate of 98.9%, efficiently delivering selected high-value components to downstream processing. It provides a reliable and scalable automated solution to the pressing challenge of EoL phone disassembly. Additionally, the automated system can enhance disassembly economics, converting a previously unprofitable process into one that yields a net profit per unit weight of EoL phones. E-waste presents a global challenge due to its rapid growth, high resource value, and the severe environmental and health risks from improper recycling and hazardous substances [1-3]. Global e-waste surged to a record 62 million tonnes in 2022 and is expected to reach 82 million tonnes by 2030 [4]. Recycling converts e-waste components into valuable raw materials, which is critical for addressing the escalating e-waste problem and supporting a sustainable circular economy [5-10]. Nevertheless, only 22.3 % of e-waste was recorded as recycled in 2022 [4]. The high human labor cost and health risk concerns are the major challenges associated with the recycling process [11]. This material is based upon work supported by the REMADE Institute, USA (21-01-RM-5083).

Rehabilitation therapy for stroke patients faces a supply shortage despite the increasing demand. To address this issue, remote monitoring systems that reduce the burden on medical staff are emerging as a viable alternative. A key component of these remote monitoring systems is Human Action Recognition (HAR) technology, which classifies actions. However, existing HAR studies have primarily focused on non-disable individuals, making them unsuitable for recognizing the actions of stroke patients. HAR research for stroke has largely concentrated on classifying relatively simple actions using machine learning rather than deep learning. In this study, we designed a system to monitor the actions of stroke patients, focusing on domiciliary upper limb Activities of Daily Living (ADL). Our system utilizes IMU (Inertial Measurement Unit) sensors and an RGB-D camera, which are the most common modalities in HAR. We directly collected a dataset through this system, investigated an appropriate preprocess and proposed a deep learning model suitable for processing multimodal data. We analyzed the collected dataset and found that the action data of stroke patients is less clustering than that of non-disabled individuals. Simultaneously, we found that the proposed model learns similar tendencies for each label in data with features that are difficult to clustering. This study suggests the possibility of expanding the deep learning model, which has learned the action features of stroke patients, to not only simple action recognition but also feedback such as assessment contributing to domiciliary rehabilitation in future research. The code presented in this study is available at https://github.com/ye-Kim/MMeViT.

Abstract--Large Language Models (LLMs) are increasingly susceptible to jailbreak attacks, which are adversarial prompts that bypass alignment constraints and induce unauthorized or harmful behaviors. These vulnerabilities undermine the safety, reliability, and trustworthiness of LLM outputs, posing critical risks in domains such as healthcare, finance, and legal compliance. In this paper, we propose GuardNet, a hierarchical filtering framework that detects and filters jailbreak prompts prior to inference. GuardNet constructs structured graphs that combine sequential links, syntactic dependencies, and attention-derived token relations to capture both linguistic structure and contextual patterns indicative of jailbreak behavior. It then applies graph neural networks at two levels: (i) a prompt-level filter that detects global adversarial prompts, and (ii) a token-level filter that pinpoints fine-grained adversarial spans. Extensive experiments across three datasets and multiple attack settings show that GuardNet substantially outperforms prior defenses. Despite its structural complexity, GuardNet maintains acceptable latency and generalizes well in cross-domain evaluations, making it a practical and robust defense against jailbreak threats in real-world LLM deployments. I. Introduction Large Language Models (LLMs) have become central to a wide range of applications, powering systems in domains such as education [1], healthcare [2], finance [3], law [4], and customer support [5]. Their ability to understand and generate human-like text has enabled automation of complex tasks such as legal reasoning, clinical triage, financial analysis, and policy drafting. However, this general-purpose capability also makes LLMs vulnerable to misuse. In particular, LLMs are highly susceptible to prompt-based adversarial attacks, especially jailbreak prompts [6], [7], which are carefully engineered inputs designed to bypass alignment constraints and elicit unauthorized or harmful responses.

Standard training metrics like loss fail to explain the emergence of complex capabilities in large language models. We take a spectral approach to investigate the geometry of learned representations across pretraining and post-training, measuring effective rank (RankMe) and eigenspectrum decay ($α$-ReQ). With OLMo (1B-7B) and Pythia (160M-12B) models, we uncover a consistent non-monotonic sequence of three geometric phases during autoregressive pretraining. The initial "warmup" phase exhibits rapid representational collapse. This is followed by an "entropy-seeking" phase, where the manifold's dimensionality expands substantially, coinciding with peak n-gram memorization. Subsequently, a "compression-seeking" phase imposes anisotropic consolidation, selectively preserving variance along dominant eigendirections while contracting others, a transition marked with significant improvement in downstream task performance. We show these phases can emerge from a fundamental interplay of cross-entropy optimization under skewed token frequencies and representational bottlenecks ($d \ll |V|$). Post-training further transforms geometry: SFT and DPO drive "entropy-seeking" dynamics to integrate specific instructional or preferential data, improving in-distribution performance while degrading out-of-distribution robustness. Conversely, RLVR induces "compression-seeking", enhancing reward alignment but reducing generation diversity.