Papers

Universal Manipulation Interface (UMI) pairs wrist-camera videos with camera trajectories and gripper openings, providing scalable, robot-free data for learning how bimanual actions change a shared scene. Existing interactive world models, however, typically rely on text or single-camera control and struggle to follow each arm's actions while maintaining consistency between the two wrist views. We introduce UMI-WM, an action-conditioned bimanual world model that jointly generates both wrist views from observation history, per-arm relative camera poses, and gripper widths, with cross-view consistency, action following, and visual fidelity. UMI-WM injects each arm's actions into its own wrist stream, while sparse hub tokens exchange scene information between the two streams. Camera-aware geometry conditioning and camera-guided noise further support per-view motion control and cross-view geometric exchange. We adapt the model for block-causal generation and distill the 30-step Teacher into a Student with four denoising steps. We train UMI-WM on self-collected open-world and in-studio UMI recordings combined with public manipulation data. To evaluate action following, we use an inverse dynamics model (IDM) to recover relative camera poses and gripper widths from generated videos, and introduce Bimanual Action Fidelity (BAF) to measure how closely these recovered actions match the conditioning actions. UMI-WM achieves the best results on all five reference-video fidelity metrics and the best or competitive performance on VBench and WorldArena. The Teacher and Student outperform all compared baselines in action following, while the four-step Student generates paired wrist videos 6.9 times faster than the Teacher.

Thin filaments cross in microscopy, and dispersed spectra cross on X-ray detectors. At a junction, where these structures cross or branch, how many straight arms meet, and which way do they go? Unknown blur obscures the answer. At a balanced crossing, distinct Gaussian blurs can even produce the same observation law. We show that the arms can still be recovered. For separated rays of nonvanishing weight and an unknown centered Gaussian-mixture blur, a distance combining arm count, directions, relative weights and supporting lines is bounded by a constant times the Euclidean difference in cell probabilities on a sufficiently fine fixed grid. The bound is uniform over a bounded model class, including unidentifiable blurs. From $n$ positions or photon counts, the minimax error is of order $n^{-1/2}$. This rate is proved for a global estimator. In practice we fit rays and blur by constrained local likelihood, reducing the BIC count when two detectors agree on fewer arms in the microscopy studies. Across three studies with protocols fixed before evaluation, on 360 crossings and branches assembled from measured single-molecule localizations, the fit gets the arm count and every direction within $10^\circ$ 24 to 37 percentage points more often than a tuned fixed-scale Hessian ridge detector, averaged over 25 to 400 localizations. On acute angles with correct counts, mean absolute error is $3.5^\circ$ to $4.3^\circ$, versus $10.4^\circ$ to $15.5^\circ$. On Chandra spectra, 38 of 90 fits at 800 photons find all four arms within $10^\circ$; among these, median error is $0.023^\circ$.

Optic disc (OD) and optic cup (OC) segmentation is central to automated glaucoma assessment, but models trained on limited labeled fundus datasets often degrade under domain shift. Semi-supervised Teacher-Student learning can improve generalization by leveraging unlabeled data, yet standard Noisy Student training treats all pseudo-labels as equally reliable. In this paper we investigate a targeted extension of Noisy Student training in which OD- and OC-specific pseudo-label supervision is modulated using a mask-derived reliability estimate. A multitask Dice predictor estimates OD- and OC-specific pseudo-label quality from Teacher-generated masks, and these estimates are used to weight the pseudo-label loss during Student training. This preserves the baseline contribution of low-scoring foreground pseudo-labels while assigning greater relative emphasis to pseudo-labels predicted to be more reliable. Experiments across known and unseen domains show improvements over the Noisy Student baseline, particularly for OC segmentation, while maintaining competitive generalization performance.

Characters, locations, and actions recur throughout a movie, so many of its moments differ only in fine-grained details. Grounding such a moment in a full-length movie requires resolving that detail rather than matching the attributes it shares with other moments, yet existing temporal movie grounding benchmarks rarely test this. We introduce CoM$^2$-Bench, a benchmark of confusable query pairs, where the two queries in each pair share most attributes but differ in a single distinguishing detail within a full movie, posing a challenge for recent temporal grounding approaches. To locate where grounding breaks down, we analyze a retrieve-then-localize pipeline, in which a retriever first selects a short candidate chunk and a localizer then performs temporal grounding within it. Separating the two stages lets us attribute each failure to either retrieval or localization. We find that retrieval is the major bottleneck: state-of-the-art retrievers often select a moment's confusable counterpart, because their choices follow the content the two moments share rather than the detail that tells them apart. To address these failures, we propose a simple, training-free remedy, Agentic Window Search (AWS): an MLLM agent $\textit{Decomposes}$ the query into attribute-level sub-queries, then $\textit{Inspects}$ each candidate window against them and $\textit{Expands}$ the window when evidence is cut off at its edge. AWS works with existing retrievers and localizers without additional training, raising mIoU from 18.5 to 26.8.

Large image editing models increasingly infer editing intent from both textual and visual inputs, creating an underexplored safety risk when harmful intent is distributed across modalities. In this work, we formalize this split-intent vulnerability and propose the Split-Intent Jailbreak Attack (SIJA), which asymmetrically distributes an unsafe editing intent by specifying the editing referent in text while concealing the safety-critical action and specification through graphical visual semantics in an auxiliary cue. To systematically study this threat, we introduce the Split-Intent Image Editing Safety Benchmark (SI2ESBench) for evaluating image editing models under split-intent inputs. Extensive experiments on representative commercial and open-source models demonstrate that SIJA effectively compromises state-of-the-art image editing systems, achieving ASRs of 68.28% on GPT Image 2 and 91.38% on Qwen-Image-3.0-Pro. Our results expose a mismatch between multimodal editing and safety assessment: the distributed intent can be recovered to execute the requested transformation, yet its harmfulness often remains undetected by existing safeguards. To mitigate this vulnerability, we propose editing-intent reconstruction, a defense principle that explicitly evaluates the complete transformation implied by the joint multimodal context and substantially improves rejection of split-intent attacks. Our findings expose a cross-modal safety gap in modern image editing systems and highlight the need to assess the jointly inferred transformation rather than individual inputs in isolation. https://github.com/yexianmin/SIJA

Vision-Language Models (VLMs) are increasingly deployed as general-purpose multimodal assistants, making multimodal safety alignment an important challenge. Safety depends on the interplay between textual and visual content, with visual context often playing a critical role. However, VLMs may fail to use visual information appropriately, either overlooking safety-relevant visual content or over-interpreting sensitive visual content, leading to unsafe responses or over-refusal. Existing multimodal safety alignment datasets largely follow the response-preference paradigm, which teaches models how to respond but provides limited supervision on how visual context should inform the safety judgment, making it difficult to balance safety and over-refusal. To address this limitation, we introduce ContraSafe-VL, a Contrastive Safety preference dataset for Vision-Language Models. For each fixed textual query, ContraSafe-VL constructs paired visual contexts that result in multimodal queries with different safety labels, providing direct supervision for how visual context should inform multimodal safety judgments. It further provides response preference pairs based on harmlessness and helpfulness to encourage models to generate responses that are both safe and helpful. Experiments across three VLM architectures and multiple benchmarks demonstrate that training with ContraSafe-VL improves cross-modal safety recognition and achieves a better safety–over-refusal trade-off while preserving general multimodal capabilities. The dataset will be publicly released.

Real-world egocentric videos capture multi-step household procedures, but their recorded trajectories leave alternative actions and recovery from the learner's own mistakes largely unexplored. We introduce EgoVine, a post-training framework that turns a fixed demonstration set into interactive experience for long-horizon planning. EgoVine operates in executable symbolic environments compiled from real-world video annotations, where an agent must coordinate object manipulations under partial observation and account for delayed consequences. Restoring a decision state allows the learner to explore alternative actions and compare their eventual outcomes. Union jointly trains action generation and consequence prediction in one model, using the same executions to supervise candidate comparison and recovery. Verified repairs add recovery targets from failed attempts, while comparisons among sampled continuations guide policy updates. The updated planner then returns to the environments, with its own decisions and mistakes shaping the next round of experience. On a shared benchmark, EgoVine-1.5B achieves 81.8% success on Main planning tasks and 66.7% on longer task compositions, outperforming all evaluated Flash LLM baselines on both. EgoVine connects real demonstrations to an iterative learning process that trains compact planners on choices and futures beyond the recorded trajectories.

AI-generated video (AIGV) detection is increasingly important for mitigating misinformation, impersonation, and other malicious uses of synthetic media, yet existing detectors often generalize poorly to unseen generators in open-world settings. To better understand this challenge, we analyze the failure modes of AIGV detectors under cross-generator out-of-distribution (OOD) shifts and observe that detector representations strongly encode generator-specific cues, which are associated with substantial ID-OOD performance drops. Further analysis shows that naively perturbing generator-sensitive tokens can even reduce OOD performance, indicating that generator-related cues and detection-relevant evidence are only partially separable. Motivated by these findings, we propose Forensic Obfuscation of geneRator-specific cues via Gradient-guided Embedding perturbation (FORGE), a token-aware adversarial training framework for open-world AIGV detection. FORGE identifies tokens that are strongly associated with generator-specific shortcuts yet less critical to real/fake discrimination, and selectively perturbs them during training to suppress generator-dependent patterns while preserving useful forensic evidence. Through synchronized adversarial training between an attacker and a detector, FORGE encourages the detector to learn more transferable generator-invariant forensic representations. Experiments on 29 held-out benchmark entries show that FORGE achieves an average F1 of 92.62, outperforming eight strong baselines by 5.31 points, while also remaining robust under four common post-processing attacks. We will release our code and datasets upon acceptance.

Real-world physics often unfolds over a variety of characteristic lengths and time scales. These non-uniformities are among the main reasons numerical simulations remain expensive when used predictively on practical systems: the same computational domain can contain smooth transport, vortical structures, thin interfaces, reaction fronts, and rapidly relaxing regions. Neural operators are attractive surrogates for such problems. However, a single architecture is still expected to represent all these local behaviours with a single set of weights and a single inductive bias. In this work, we introduce the regime-aware coupled neural orchestrator (RCNO), a framework that learns to distribute and connect a library of local operator specialists in space and time. For every patch, an orchestrator builds a causal query from the current fields, neighbouring halo values, local regime features, patch identity, activity level, and case parameters. The nine expert embeddings exchange two rounds of query-conditioned message passing before an edge-biased cross-attention stage produces a dispatch distribution that can be evaluated sparsely. Local predictions are assembled into a global field and projected onto system-specific physical constraints. A seam-aware extension additionally learns residual corrections across internal patch interfaces. We apply the same orchestration architecture to two qualitatively different systems: reacting flow and Cahn-Hilliard-Navier-Stokes (CHNS) dynamics. The results show that useful operator specialisation exists across local regimes, that the advantage of conditional composition increases as the spatial decomposition becomes finer, and, on reacting flow, that explicitly treating patch interfaces improves continuity, PDE consistency, and rollout accuracy.

Retrieval-augmented generation asks one question at volume: does this evidence support this claim? The answer is one bit, yet most pipelines obtain it from a text generator and pay generation prices for the prose wrapped around it. We compare a non-generative decision model, which returns a typed value from a declared enumeration together with a native probability and no text, against a frontier LLM configured as an engineer would deploy it for binary classification. Both answer the same 495 human-labelled claims from LLM-AggreFact, so every comparison is within-item. We report three things. First, the two systems are statistically indistinguishable on the metric this task is conventionally scored by, a paired difference of -0.56 points, Jev minus Astra, at a 187-fold cost difference. Second, that tie carries almost no information about deployment: the systems disagree on 15.2% of claims, and the cheaper one admits 23.2% of the claims annotators marked unsupported against the frontier model's 14.1%. A component whose job is to catch unsupported claims should be judged on that rate, and the conventional headline hides it. Third, because the errors differ they compose: routing the claims the cheap model is confident about and escalating the rest reaches 74.10 at USD 2.92 per thousand claims against 73.83 at USD 8.93, with the threshold chosen out of fold. But a matched random-escalation control at in-sample thresholds, spending the same budget by coin flip, does as well, so what buys the saving is the escalation share and not the probability that picks what to escalate. Measured against the cheap model it is built from, the cascade costs 61-fold more for 0.83 points, 95% CI [-3.36, 4.86], and it is the more permissive guard of the two. Equal accuracy conceals unequal errors. All predictions, prompts, pricing and code are released.

Long-horizon failures of language models are usually measured as end-task success or as a horizon length. We instead score, step by step, three components of a 400-step ledger task in which every step is easy in isolation and the model's own answers are re-injected into its context: recognising a record already seen 5 to 300 steps earlier, processing the current record locally, and writing a counter incremented at every step. On Qwen3-8B (non-thinking, 20 seeds, replicated on a second machine and backend), with every step scored component by component, two results hold. First, duplicate-recognition errors rise late in the episode – from 0.0019 to 0.011 per step at risk across step 200 – while the counter makes no error in 8,000 steps and local processing is descriptively stable; the rise combines long-distance recall errors and rising false alarms. Second, a block of irrelevant content inserted before each step makes the counter fail in 20 of 20 seeds on each machine, against none without it, while its effect on local processing is much smaller and not distinguishable from none. Three pre-registered follow-ups probe the counter. Its value depends on the value present in the response history: a corrupted value persists without correction in every seed, and an exact external reminder, repeated before each message, re-aligns it within one to two steps. Under the irrelevant block, no counter error was observed when an exact compact reminder of the state, supplied externally by an oracle, preceded each message (0 of 20 seeds with an error, against 15 without it), and this reminder outperformed a length-matched neutral text; whether a single reminder suffices was not conclusive, and the absence of damage to the other components was not established. An aggregate or exclusive error classification can hide the difference we report.

Modern healthcare generates increasingly complex medical reports that challenge clinical professionals in terms of time and interpretive burden. Manual review processes are slow, error-prone, and difficult to scale. This paper presents MeddeX, a Medical Report Analyzer that combines a QLoRA fine-tuned LLaMA-3 (8B) model with Retrieval-Augmented Gen eration (RAG) and a multimodal Gemini Vision layer to deliver accurate, concise, and evidence-backed summaries of medical documents. The system handles text-native PDFs, scanned image based PDFs, and direct radiological images through an intelligent hybrid extraction pipeline that includes a novel garbage-text detection mechanism for corrupted and UTF-16-encoded doc uments. Beyond single-report Q&A, MeddeX introduces report scoped retrieval for precise contextual grounding, an automated longitudinal report comparison engine, and a self-healing backfill pipeline that lazily processes legacy documents on demand. A FastAPI backend, Supabase vector store, and Next.js fron tend complete the end-to-end architecture. Evaluation against the base LLaMA3-Med42-8B model demonstrates consistent improvements across ROUGE-1 (+9.2%), ROUGE-2 (+10.0%), ROUGE-L (+11.8%), BLEU (+71.4%), and BERTScore (+0.8%), validating the benefit of domain-specific fine-tuning. MeddeX represents a practical and robust step toward AI-assisted clinical decision support.

Romance scam has become a serious problem via social media, and the improvement of artificial intelligence enable to fake identify easily. In the study, a chatbot has developed to detect romance scam in real-time through Line Messaging API. Using the seven stage theories in the social science field, the stages of romance scam can be detected through large language model. In addition, once a romance scam is detected, the chatbot provides explanations of the stage and prevention advices through large language model. The contribution of the work provides a practical application, and can be used in real scenario for reducing the risk of romance scam.

Modern data pipelines increasingly combine human-created and human-curated data with AI-generated data for model training and evaluation. As datasets grow, they often add semantically redundant samples rather than expanding semantic coverage, which deteriorates the quality of the whole dataset. Existing diversity metrics primarily summarize lexical variation, entropy, or pairwise similarity, but provide limited characterization of the shape of the underlying data distribution. In this paper, we propose PLDiv, a diversity metric derived from persistence landscapes that quantifies the structural organization of a dataset from its geometric representation. PLDiv summarizes how data neighborhoods evolve as nearby samples progressively connect, producing an interpretable scalar measure of structural diversity without requiring task-specific supervision. Whether measuring intrinsic diversity in text generations or image embeddings, or detecting semantic collapse in instruction datasets and reasoning trajectories, PLDiv performs competitively against evaluated baselines and aligns strongly with diversity labels during semantic collapse. Consequently, it effectively supports applications in synthetic data monitoring, dataset construction, data augmentation, and generative model evaluation.

Looped transformers reuse one shared block across many recurrent steps, so depth becomes a quantity that can be spent at runtime rather than a stack of distinct layers. Those extra steps help only if a write made on one recurrent step is still readable on later ones, and that depends on the linear map applied to the residual between recurrent steps. We show that whether the loop learns to track state is a geometric question about the form of this map, and we propose blockwise conformal injections: the residual splits into coordinate-aligned groups and rotates inside each one, so every group has its own timescale while the geometry stays exact within it. Our experiments show a performance increase on a demanding form of state tracking - last-token accuracy when the rest of the prefix is already correct - and on language modelling, where a signal written earlier has to be retrieved at the end of the sequence.

LLM-as-a-judge evaluation commonly aggregates multiple judges' decisions into a majority vote, yet this panel decision process does not guarantee that the evaluation of an individual item is reliable. We propose Structured Cross-fitted Uncertainty Risk (SCUR), a label-free approach for estimating instance-level evaluation reliability based on the judge panel's behavior. A latent measurement model serves as the backbone for measurement system analysis, from which SCUR derives three complementary signals capturing panel uncertainty, measurement cross-entropy, and perturbation sensitivity. These three sources of evidence are combined into a risk score. We evaluate the SCUR approach on five real-world benchmarks using majority-vote error ranking, selective trust, and risk stratification. Furthermore, we show our approach to be more effective at detecting high-consensus risk items, which are the blind spots for many other LLM-as-a-judge evaluation methods.

Data selection for LLM fine-tuning faces two facts about real training: data pools draw from many distinct domains, and evaluation spans diverse tasks. As a result, the same sample helps different tasks differently, and samples interact during training so that a subset's value is not the sum of its parts. Current data-selection methods leave these realities largely unaddressed. We formalize them as Combinatorial Data Selection (CDS), a framework that separates attribution (how samples are scored) from selection (which subset to pick) and models the non-additive interactions among samples. On a real instruction-tuning data pool, we find substantial cross-domain complementarity (joint utility exceeds the sum of singletons), alongside redundancy within domains. Inspired by this, we propose Mixture-then-Select, a two-stage algorithm that first decides how many samples to take from each domain, then ranks samples within each domain by attribution score. Theoretically, the gap to the optimal subset is linear in the selection size and independent of data pool size. Empirically, Mixture-then-Select outperforms strong baselines by 20 to 54%, is robust across attribution methods, and works across different models and data pool sizes. The framework also extends to the data purchasing problem under a budget constraint when sample prices vary. Our findings suggest that allocating selection across domains offers a promising path toward more efficient and effective data selection for modern LLMs.

SIREN (Ship Inviscid Resistance and Elevation eNgine) is a Rankine-source panel solver for the steady wave resistance of a ship, in the tradition of Dawson's method. The free-surface condition is linearised about a base flow, and the linearisation is iterated. In the present formulation the base flow is the solution of a Neumann problem on the actual boundary: sources on the entire hull, wet and dry, and on the free surface panels, with impermeability imposed on both, so that the free surface acts as a rigid lid at the position it currently occupies. This replaces the classical double-body flow, whose image about z=0 is meaningful only for a flat surface and a hull cut at z=0, and it removes the need for desingularised free-surface panels. In the non-linear iteration the free-surface panels are moved rigidly in the vertical to the relaxed wave elevation at their centroid, the hull panels are marked wet or dry against the moved surface, and the base flow is recomputed on the moved geometry; impermeability applies to the whole hull regardless of this marking, which is used only to select which panels enter the force integral, and under-relaxation is confined to the passes where the non-linear correction accumulates without being reset. The radiation condition is imposed by a one-row upstream shift of the collocation points, and the wave resistance is obtained by pressure integration and, independently, by transverse wave-cut analysis.

The parity problem in 1D Cellular Automata is a classifier based on a local rule. If the number of ones in a ring of binary cells is odd then the automaton shall converge to all-one (11...11) else to all-zero (00...00). The objective is to find rules that are related to Fatès's Balanced Rule F but use less active neighborhoods (actions). The method is to analyze the effect of each actions and to find a functioning subset of the 12 actions used in rule F. New rules called HA, HB, HC, HF are defined with 10, 6, 4, 8 actions. Rule HC and HF need only three inputs whereas all others need four. The classification time of the rules was determined by averaging over many simulation runs with random initial configurations. It is almost the same for the rules HA and F, and is longer for the rules HB, HC, HF. In addition, the rule HD is defined, which updates three cells in parallel and works faster. In future work, the functioning and performance of the proposed rules should be treated in a more formal way, and related rules (also with different update schemes) could be considered.

Many headphone users risk irreversible hearing damage during high-load listening, yet current protection still relies on sound-pressure limits. Clinical tests detect damage after it becomes measurable, leaving the risk-accumulating process unobserved. The acoustic reflex (AR), an involuntary response to high acoustic load, marks when the ear enters a protective state. AR traces are weak, and existing measurements require dedicated probes and controlled conditions that are difficult to maintain during everyday audio. We present EarLog, which turns playback, the dominant interference, into an opportunistic probe. EarLog introduces content-conditioned residual sensing to suppress variations caused by nonstationary content, individual ear acoustics, and device response. It identifies weak AR through loudness-dependent changes and localized spectral irregularities in the residual. A resonance-constrained sparse finite impulse response (FIR) model captures the short-window response. Dual-microphone observations, mild-distortion recovery, and cross-session reference updates keep detection usable under low signal-to-noise ratio (SNR), fit changes, and environmental drift. EarLog runs in real time on an embedded controller using a low-cost dual-microphone add-on. We evaluate it with 81 participants, 16 headphone models, and more than 400 hours of recordings. EarLog achieves 99.0% sensitivity and 98.7% specificity on a clinically referenced benchmark. EarLog records listener-specific AR events during everyday listening, providing data for future longitudinal studies of hearing risk.

Alzheimer's disease (AD) remains a major public health challenge, and early identification is crucial for timely intervention and care planning. Recent work has shown that speech-based digital biomarkers can capture subtle changes in vocal production, prosody, and timing that often precede or accompany cognitive decline. In this study, we investigate whether acoustic features extracted from spontaneous speech can reliably distinguish individuals with AD from healthy controls. Using a structured speech dataset derived from the Cookie Theft picture-description task, we extract clinically informative low-level descriptors spanning prosodic, spectral, and temporal domains, and evaluate a range of machine learning classifiers. Our results indicate that acoustic biomarkers carry strong discriminative value, with the best-performing models achieving clinically meaningful separation between diagnostic groups. These findings support the use of non-invasive, low-cost audio-based screening tools and motivate further work on multimodal and longitudinal assessment frameworks for dementia risk detection.

We systematically explore the relationship between edge-connectivity and Lin–Lu–Yau curvature for graphs. The intuition is that a connected graph with large Lin–Lu–Yau curvature should also have large edge-connectivity, and vice versa under suitable conditions. We prove that the edge-connectivity of a connected graph is bounded lower by the product of its minimum degree, its Lin–Lu–Yau curvature and the constant $\frac{3}{2}$. We also provide two lower bounds on the edge-connectivity of a graph that guarantee positive Lin–Lu–Yau curvature, and discuss the sharpness of these bounds.

Proton therapy offers a dosimetric advantage over conventional radiotherapy through the Bragg peak, enabling precise dose delivery to the tumour while sparing surrounding tissue. Monte Carlo (MC) simulation provides the most accurate dose calculation but is computationally expensive for clinical workflows. A promising compromise is to simulate with a reduced number of primary histories and recover a high-quality dose distribution through a deep learning denoising step. In this work, we take a 3D multiscale U-Net from recent literature as a case study and investigate two improvements critical for clinical deployment. First, we replace the standard MSE training objective with a differentiable $\gamma$-index-based loss, extended to native 3D operation, and demonstrate that directly optimising the clinical evaluation metric yields a mean Gamma Passing Rate of 99.47% at 2mm/2% and 95.67% at 1mm/1% on the test set, compared to 92.10% and 71.13% for MSE. Second, we apply L1-norm structured filter pruning to reduce the model's memory footprint, achieving a 13.3% parameter reduction with negligible performance loss under partial pruning, and 20% reduction under full-block pruning after fine-tuning.

Identifying suicide risk from social networking services (SNS) posts is important for detecting suicide-related signals in online environments. However, risk classification alone provides limited insight into the textual evidence and psychosocial factors behind a prediction. Based on the IEEE BigData 2026 Explainable Suicide Risk Detection Challenge, this study presents a framework consisting of Risk Assessment, Evidence Grounding, and Factor Identification. Risk Assessment uses length-based routing to accommodate posts of different lengths. Evidence Grounding identifies supporting phrases and uses a Risk-Evidence constraint to maintain consistency with the Risk prediction. For Factor Identification, two verifiers are used. The Taxonomy Verifier focuses on factor semantics, whereas the Evidence-Aware Verifier uses factor-specific lexical-semantic cues to select informative positive training units. Their prediction probabilities are combined to produce the final factor predictions. The three tasks are evaluated using task-specific F1 score measures. Risk Assessment achieved a Weighted F1 of 0.8088, Evidence Grounding achieved a test Macro row F1 of 0.7605, and Factor Identification achieved a Macro F1 of 0.5562. The results show that the framework can provide risk predictions, along with supporting textual evidence and fine-grained information on psychosocial factors. Overall, the proposed framework extends suicide-risk assessment beyond risk-level prediction and provides a more interpretable analysis of SNS posts.

Quantum modifications of black hole spacetimes can lead to observable departures from the predictions of classical general relativity, particularly in the strong-gravity regime near the event horizon. In this context, the coherent quantum black hole model introduces a sine-integral correction to the gravitational potential, characterized by a coherence scale \(R_s\), which softens the central singularity and alters the near-horizon geometry. Motivated by the growing precision of black hole imaging and strong-field observations, we explore how these quantum-induced effects influence photon motion and the resulting shadow structure. Considering representative values within the interval \(R_s \in [0.5M, 1.2M]\), we show that the photon sphere shifts inward while the critical impact parameter decreases, reaching values such as \(b_{\rm crit}\approx 4.764M\) for \(R_s=0.5M\). Consequently, the shadow radius becomes systematically smaller and develops a more intricate structure than in the Schwarzschild case (\(b_{\rm crit}=3\sqrt{3}M \approx 5.196M\)). The analysis of the corresponding intensity distributions also reveals a pronounced enhancement of radiation near the shadow boundary, especially for intermediate coherence scales. This feature can be attributed to photons remaining longer within the modified near-horizon region before escaping toward distant observers. Such an increase in brightness may therefore provide a possible observational signature of quantum corrections at horizon scales. Furthermore, although the geometry allows radiation and energy to accumulate around an effective horizon surface, its response to perturbations differs from that expected for compact objects with stable light rings. Instead of contracting under backreaction effects, the accumulation surface shifts outward.

Large language models (LLMs) often abandon a correct answer, or endorse a user's position, once the user pushes back. This behavior, called sycophancy, is usually reported as a single rate per model, which says little about when it happens or how a user can avoid it. We study the conditions that produce it with 103,939 graded replies from ten configurations: eight LLMs with reasoning disabled, and two of them again with maximum reasoning, all facing the same 200 items, 13 pressure conditions, and four-turn conversations, with every reply labeled by two independent LLM judges. We find that the dominant factors are how costly it is for the model to verify the user's claim, and whether a trained guardrail covers it. Removing this task factor from a logistic model costs 0.485 of McFadden $R^2$, against 0.139 for model family and 0.009 for pressure tactic. Anchored facts are almost never conceded (1.3%), while adoption on logic puzzles rises with the number of clues needed to refute the pushed answer. Personal choices are endorsed in 77.0% of conversations. Most concessions on hard items come from models that cannot reliably solve them; models that can solve them rarely give the answer up. For both models tested, maximum reasoning removes these concessions completely: adoption on deep puzzles falls from 19.2% and 12.5% to 0%. Fallacious or emotional framing adds nothing beyond plain repetition. Three human annotators agree with the judges' consensus on 118/120 calibration items. These results give practical rules for reliable use: simplify hard-to-verify problems and reason deeply, state the question rather than one's preferred answer, ask for evidence on open questions, and choose models by their measured guardrail profile.

Language training relies on a corpus constructed by a large number linguistic materials. AI-powered voice clones provide a way to construct the corpus with relatively low cost. Singing voice conversion (SVC) model is used to generate synthetic voices. This study compares participants' performances on natural and synthetic speech in two experiments, similarity perception and intonation recognition. In the accuracy of similarity perception task, a significant interaction between speech type and intonation is found, suggesting that question may serve as a cue for speaker identification but may be influenced by synthetic features. In the accuracy of intonation recognition task, a significant interaction between speech type and familiarity is observed, indicating that speech type affects how much familiarity contributes to voice processing.

Activation steering manipulates large language model behavior by intervening on internal activations, but the mechanistic basis of these interventions remains poorly understood. We decompose refusal steering into component-level interventions across four open-weight models, identifying the sparse subsets of attention and MLP components whose steering suffices to reproduce the full behavioral effect. We find that refusal directions concentrate in sparse component mechanisms comprising 28–48% of upstream components, retaining 88–101% of steering effectiveness. Within these mechanisms, effective steering further concentrates in approximately 50% of residual stream dimensions, retaining 85–98% of the component-mechanism baseline, consistent with a privileged basis structure. Sparsity thus operates at two levels: which components are steered, and which dimensions within those components carry the signal. Together these findings show that refusal is not diffusely encoded across a transformer but assembled by a structured, identifiable mechanism, providing a foundation for mechanistic understanding of how refusal behaviors are represented and steered. To facilitate reproducibility, we release all code and raw experimental results in https://github.com/wang-research-lab/Refusal_Mechanisms.

Retrieval-augmented generation grounds language models in external context, but for long documents flat top-$k$ retrieval can cluster on a single region and miss complementary evidence. RAPTOR-style summary trees address this by recursively clustering chunks and using a language model to summarize each cluster at indexing time, then ranking summary nodes alongside raw chunks at query time. We show the main benefit of summary trees in long-document QA can come from navigation rather than the generated summary content. We introduce NavTree, a leaves-only retriever that builds a deterministic balanced segment tree over chunks (zero language-model calls at indexing) and uses the tree purely as a navigation scaffold: a hybrid lexical-and-dense frontier walk, anchored on top retrieved leaves, descends from the root and emits only leaf chunks to the reader. On a matched-cost evaluation against flat retrievers and an extractive re-implementation of RAPTOR, NavTree is the strongest matched-cost hierarchical retriever in our evaluated grid and ties the strongest flat baseline. On long-document multi-hop QA, it is the only hierarchical method that significantly beats BM25 on a class-vs-class basis, corroborated by a reader-free retrieval-recall check. A matched-reader replication of the published abstractive RAPTOR variant, given strong cluster summaries, still loses to NavTree at every multi-chunk budget, at zero indexing cost. The ranking carries across stronger and open-weight readers, a stronger encoder, and a full factorial that isolates leaves-only emission as the structural lever.

In this communication, we discuss results obtained with the numerical in-house PHP simulation code CASCO (French abbreviation for Advanced PHP simulation code). The aim is to study the impact of bubble generation (boiling) on the established oscillation of a PHP with a high turn number (10-turn water-copper tubular PHP with independently heated turns in the evaporator region). We vary the nucleation barrier value, i.e. the local superheating $\Delta$Tnucl at which the nucleation occurs and the heat load Pe; other simulation parameters are kept constant. The PHP heat transfer performance is evaluated with the PHP thermal resistance Rth for both horizontal and vertical favorable PHP orientation. Since the bubble generation plays a positive role for the PHP functioning, one expects that the function Rth($\Delta$Tnucl) monotonously increases. This is true for high $\Delta$Tnucl. For small $\Delta$Tnucl, the Rth($\Delta$Tnucl) function is decreasing, thus a minimum occurs. It appears because of a transition between the regimes of stable PHP functioning. The regimes strongly differ by the average penetration of liquid plugs into the evaporator. For horizontal PHP, these regimes correspond to the intermittent and continuous oscillations. The regimes are discussed and the regime map in the coordinates Pe -$\Delta$Tnucl is obtained.