Existing methods mainly rely on convolutional aggregation or attention-based spatial-spectral interaction. However, conventional feature-domain attention does not explicitly characterize variation patterns along spatial-token and learned latent-channel coordinates, while spatially shared local operators lack the flexibility to adapt neighborhood aggregation to heterogeneous structures. To address these limitations, we propose a Frequency-Structure Complementary Modeling Network (FSCM-Net), which couples frequency-aware dependency learning with structure-adaptive local refinement. Specifically, a Spatial-Spectral Frequency Calibration Attention (SSFCA) module constructs spatial-token affinities from Fourier-interacted query-key representations and builds latent-channel affinities after input-conditioned frequency filtering along learned channel coordinates. The latent-channel response is then calibrated by a learned shared gain and combined with the spatial-token response. Complementarily, a Structure-Coherent Dynamic Feature Fusion (SC-DFF) module combines regularized structure-tensor cues from MSI features with context-modulated HSI detail representations to generate channel- and position-dependent aggregation kernels. This design makes local HSI refinement responsive to smooth regions, edges, and textures while maintaining spectral fidelity. The two modules are progressively integrated within a multiscale encoder-decoder to coordinate representation-level interaction and structure-adaptive reconstruction. Experiments on two simulated benchmarks and one real remote-sensing dataset demonstrate favorable performance against representative fusion methods in spatial detail reconstruction and spectral preservation.
Large Language Models (LLMs) are increasingly deployed through conversational interfaces, yet strong single-turn performance does not imply reliability in multi-turn settings. Multi-turn assistance requires a model to maintain an evolving representation of user requirements and to reason over it before complying with later requests. Failures can therefore be subtle: models may fail to recognize a changed constraint or hidden conflict, or recognize it without enforcing it in the final response. Existing benchmarks primarily test fully specified prompts, long-context recall, or controlled domain environments, leaving open-domain requirement reconciliation undermeasured. We introduce TurnFlow, an interactive benchmark for evaluating whether models preserve, revise, and act on user requirements across turns. TurnFlow combines adaptive user simulation with verifiable judging, allowing conversations to unfold naturally while keeping hidden requirements and pass/fail criteria controlled. In our evaluation, nine LLMs achieve overall pass rates from 42.86% to 59.59% on 490 scenarios. Performance varies substantially across challenge types: models perform well on Stateful Multi-Constraint Following but remain considerably less reliable on Scope Creep and Goal Shift. These results reveal persistent limitations in reasoning over evolving user requirements across turns.
LLM-based agents extend LLMs with tool calls, memory, planning, and retrieval to act in external environments, but they are vulnerable to attacks at different workflow steps. Existing defenses, including prompt engineering and post-training alignment, primarily improve LLM-only safety and often overlook risks that arise from multi-step execution. External verification adds guardrails to check each step separately, but it typically ignores how risks can propagate across steps. We argue that we should consider risk transitions across dependent steps through the agent workflow, rather than separate verification on a single step. In this paper, we propose \textbf{T}ail-risk-aware \textbf{A}gent \textbf{S}afety (TAS), a tail-risk-aware safety framework that models agent workflow safety as a sequence of dependent safety states and constrains the rare but extreme risks (i.e., tail risk) with a shared budget. TAS first models safety-state transitions sequentially along the workflow as an MDP to trace risk propagation, then measures a tail-sensitive risk to target rare but high-cost risks (i.e., tail risk), and constrains the risk with a shared budget to derive a step-wise safety strategy. By constraining tail risk rather than the overall average, TAS suppresses rare high-cost failures while preserving utility in normal cases. Experiments show that TAS achieves better performance compared with eleven baselines across three benchmarks.
List replicability measures how many accurate hypotheses suffice to contain the output of a learning rule with arbitrarily high confidence. We prove that every nonempty class of $m$ binary concepts on $n$ domain points has list replicability at most $805\ln(2m)\ln(2n)$. This resolves the logarithmic-product conjecture of Blondal, Hatami, Hatami, Lalov, and Tretiak. Hanner matrices give lower bounds of the same order, for square matrices and for explicit families with unequal logarithmic aspect ratios. The upper bound is obtained by constructing nested acyclic sets in the faces of a cube. A critical-point estimate supplies highly connected regions, weighted cell attachments keep the sets compatible across faces, and a cubical carrier theorem converts them into accurate finite covers and learning rules. Appending one constant coordinate turns list replicability into an exact geometric quantity, the face width, and graph-orientation classes satisfy exact product formulas. In particular, there are explicit classes with list replicability $4k+1$ and equivariant index $3k$, so the additive gap between the two is unbounded. We also develop entropy-tilting and two-sided Gibbs constructions, give combinatorial upper bounds adapted to structured matrices, compute exact values on classical families, and study the index of Paley matrices.
Longitudinal evidence about the psychological impact of conversational AI use remains scarce. We report a survey study of 114 adults who reported their level of problematic AI use one year after reporting their motivations (for escape, social, entertainment, and instrumental reasons) for engaging with AI. We found high levels of AI motivation of all types predicted later problematic AI use, especially when participants turned to AI for managing their feelings and overcoming difficulties. Our findings imply that AI motivation could be used as an early warning signal of potential problematic use, especially if individuals engage with AI for emotional purposes over time.
Frontier agentic LLMs are trained on verifiable rewards in reinforcement learning (RL) environments built from real software repositories. A passing test suite is treated as the reward for solving the task. We show that, on hard problems, frontier models often satisfy these rewards by obtaining the answer instead of deriving it, known as reward hacking. Prior reports document this behavior on public benchmarks. We test its persistence on unseen problems. In a hardened environment where the answer cannot leak, nearly every model we tested attempted to recover the answer, escalating through tiers of increasingly specialized exploitation: mining the repository history for the upstream fix, inspecting the git object store directly, and sweeping the filesystem for a possible newer copy of the source or hidden tests. These attempts were concentrated on the hardest problems, and the models differed in whether, how often, and how deeply they escalated. Our position is that this behavior is a learned shortcut, acquired in RL environments that leak the answer, not a sign of intelligence. We release agent traces for these models to support further study at https://huggingface.co/datasets/logitly-ai/reward-hacking-raw-v1.
Big techs and researchers increasingly claim that their training datasets are too large to inspect for bias and ethical problems. However, in the race to build ever-large models, they have relied on automated web scraping and no human curation. Datasets built this way absorb the stereotypes, exclusionary norms, and historical prejudices of their sources, and the models trained on them can reproduce and amplify those patterns. Because manually auditing billions of samples is humanly impossible, we need practical, effective methods to identify biases in these datasets. This paper studies the feasibility of ethically auditing large datasets through small samples and proposes a practical method for identifying structural biases by applying AI Ethics Tools (AIETs) to random samples. We applied Datasheets for Datasets, Data Statements, Data Nutrition Labels, and Data Cards to a 1% sample of Framed Multi30K, and Datasheets for Datasets to two 0.00006% samples of relaion2B-en-research-safe. Our results show that evaluating small samples with AIETs can reveal ethical problems, structural biases, and technical inconsistencies. Although industry practice assumes that identifying problems requires analyzing the entire dataset, we show that a random subset supports a viable qualitative analysis without overwhelming computational or human resources. Demonstrating that a dataset is biased does not require examining every instance: if a bias appears in a random sample, it is a flaw of the creation process. The AIETs' questions acted as provocateurs, evoking different perspectives and analysis methods for the same dataset, but the analysis and its results depended fundamentally on the human factor. We hope these findings contribute to urgent public discussions about ethics, transparency, accountability, and AI regulation by offering a practical pathway for responsibly auditing massive datasets.
Recent work has shown that hidden biases can be transferred through data with no apparent semantic relationship to those biases. Can regularization limit this covert influence? By constraining adaptation, regularization might be expected to suppress behavioral changes on seemingly unrelated inputs. We systematically study this hypothesis in subliminal learning, as well as in related settings of covert influence, including phantom transfer and logit-linear selection. We find the opposite: regularization amplifies the transfer of hidden biases. In particular, subliminal learning is typically absent under full fine-tuning, but low-rank adaptation or explicit penalties can induce strong transfer. Phantom transfer and logit-linear selection show similar trends. Adding more data, an implicit form of regularization, also strengthens covert influence. Together, these findings suggest that the simplicity biases induced by regularization may facilitate covert influence. We further hypothesize that the level of regularization that maximizes covert transfer depends on task complexity.
Information-flow control (IFC) is a promising approach to enforce policy in agentic AI systems. However, IFC suffers from label creep, a phenomenon where monotonically increasing security labels propagate through a system and progressively limit its utility. Label creep in agentic AI systems can be mitigated using two major strategies: (1) improved system design and (2) label declassification. In this paper, we characterize how far system design alone can mitigate label creep in agentic systems. Specifically, we measure how system topology, scale, and agent composition affect utility loss and security enforcement cost proxies for conservative IFC enforcement in a stateful agentic system simulator. We develop a pipeline to systematically vary these features and measure their impacts on label propagation. Our results demonstrate strong associations between structural graph features and our IFC utility loss metric (up to a Spearman correlation $|\rho|$ of $0.96$) and show that system size and tool allocation substantially affect how labels propagate. These findings identify design choices to reduce IFC utility loss, but they also demonstrate that system design alone cannot fully prevent utility loss due to monotonic label propagation. This motivates the development of label declassification mechanisms that aid these design choices and diminish IFC utility cost.
Wrapping C code in a safe Rust interface lets Rust applications use it without writing unsafe code, until the code itself is migrated to Rust. Such bindings are tedious and error-prone to write, however: their internals use unsafe code that the Rust compiler does not check, so a single mistake can introduce undefined behavior (UB), and they have so far been written mostly by hand. We present Crustify, an agentic co-translator that generates safe Rust bindings for repository-level C libraries. Crustify schedules a target's types and symbols into dependency-ordered waves of independent batches that agents bind concurrently. Its agents derive the ownership and lifecycle contract of every pointer and type they wrap, check each binding with coverage-guided equivalence and UB tests under sanitizers and Miri, and review every wave. They build on three tools that we contribute: (a) ffibox, Rust primitives for C/Rust interop that mirror Box, Arc, and Vec for C objects without ever forming a reference to memory that C may alias; (b) Wavefront, a neuro-symbolic static analyzer that orders a target's items by their dependencies, including those that casts and type-erased pointers hide; and (c) ffimap, a rustc-based analysis that maps the unsafe surface of bindings. On libgit2 and ffmpeg, with one developer in the loop, Crustify generated safe bindings for 41.9% and 42.8% of their public APIs in about 51.9 and 58.9 hours. On the interface consumers are expected to use, only 1.6% and 0.4% of the bindings' public functions are unsafe, and only 0.1% and 0.6% have a raw pointer in their signature. Their tests cover 98.92% and 99.07% of the Rust code and 45.95% and 67.20% of libgit2 and of ffmpeg's libavutil, and the reviewers fixed 8 and 13 defects that caused UB reachable from Rust. To our knowledge, this is the first work to automate the generation of safe Rust bindings for C libraries.
Estimating the principal eigenpair of a positive operator from stochastic evaluations gives rise to a noisy form of normalized power iteration: each step applies a sampled operator estimate and then renormalizes. We establish finite-time stochastic-approximation guarantees for this procedure, working in Hilbert's projective metric, which is well-suited to the multiplicative geometry of Perron--Frobenius eigenproblems. Our main application is risk-sensitive average-cost reinforcement learning (RL) with exponential utility. In this setting the Bellman equations are multiplicative, and value functions are characterized by nonlinear eigenvalue problems rather than additive fixed points. We cast both policy evaluation and control in this framework, obtaining risk-sensitive TD- and Q-learning algorithms that learn from a single Markovian trajectory. Under explicit positivity, mixing, and linear function approximation assumptions, we prove finite-time bounds on the recovered eigenvector and eigenvalue with $\tilde{O}(\epsilon^{-2})$ sample complexity up to problem-dependent constants. The results cover both tabular and linear function approximation regimes, and through the duality between exponential utility and KL-robust control yield finite-time guarantees for KL-robust average-cost RL as well.
World models enable planning through internal simulation, while natural language provides intuitive task specification. Yet, language-conditioned world modeling remains fragmented across domains with disparate action spaces, datasets, and evaluation protocols. We introduce LCWM, a shared task that maps an initial agent-view observation and a language instruction to an action sequence that fulfills the instruction. Its accompanying benchmark, LCWM-Bench, integrates 15 datasets across navigation, robotics, 3D gaming, and driving under a shared action interface and evaluation protocol, pairing 142,987 episodes with 428,961 instructions in three styles. We further propose Tiresias, an agent that plans via selective imagination: a language-conditioned diffusion policy generates candidate trajectories, a trajectory ranker shortlists them, and a video-pretrained world model imagines their visual consequences. A visual ranker evaluates these rollouts, complementing trajectory scores to select the final plan. Tiresias achieves competitive planning and imagination performance on LCWM-Bench. Ablations demonstrate the complementary benefits of trajectory ranking and visual verification, while experiments with alternative proposal methods show that the trained rankers remain effective beyond our diffusion policy. Together, LCWM-Bench and Tiresias provide a unified evaluation setting and an effective starting point for studying language-guided planning through world models.
When disasters strike, they not only damage infrastructure, buildings, and the economy, but also affect the mental health of affected populations. Research on the mental health effects of disasters has expanded rapidly and has become an important, well-funded area of psychology, public health, and related social sciences. However, the overall structure of this research landscape has not been systematically characterized. Using large language models, we conducted a review of all articles related to terms 'disaster' and 'mental health' indexed in PubMed. 7,859 articles' metadata were retrieved using the NCBI API, and then Claude Sonnet was used to extract structured variables from the abstracts. The results indicated that earthquakes and epidemics were the most commonly studied disasters while many other disasters such as heat and cold waves have received less attention. PTSD, depression, and anxiety were well investigated but many other mental health issues (e.g., aggression, suicide) are relatively ignored. Despite enjoying relatively large sample sizes (median = 452, IQR = 142 - 1,568), the field rarely utilizes machine learning algorithms. Hypothesis significance testing, reviews, qualitative methods, and structural equation modeling are respectively most common methods while agent-based modeling, machine learning, casual inference, and NLP each appear in less than 1% of articles. Studies are limited to a small number of countries, particularly the United States, Japan, and China, while many disaster-prone regions remain substantially underrepresented. This LLM-assisted computational map provides a broad overview of the disaster mental health literature while highlighting important methodological gaps, with particular emphasis on the underutilization of computational methods that could help advance the field.
Supervised fine-tuning over a large pool of heterogeneous tasks is often computationally expensive, leading practitioners to train language models on a fixed training budget of carefully chosen SFT tasks. Thus, achieving strong downstream performance hinges on selecting an effective subset of tasks, or equivalently, identifying an optimal task mixture under a fixed budget constraint. In practice, supervised finetuning mixtures are often determined using simple heuristics (e.g., uniform or size-proportional sampling) that ignore task interactions, which can hurt transfer and waste budget on redundant task sources. We introduce TASKMIXPGM, a framework for learning continuous task mixtures through energy-based modelling over SFT tasks. Each task forms a node in a Markov random field, with unary potentials capturing task utility and pairwise potentials modelling inter-task relationships via behavioral divergences between predictive distributions of single-task fine-tuned models. Optimizing this objective yields task mixtures that balance utility, coverage, and redundancy. We further show that the induced set function defined over the set of such tasks is weakly submodular under budget constraints, enabling approximation guarantees for discrete task selection. Empirically, we provide a fixed, benchmark-independent mixture that improves over heuristic mixing across multiple model families and different training budgets. TASKMIXPGM also uncovers interpretable structure in task interactions, revealing complementarities and redundancies that inform effective SFT mixtures.
Standard causal discovery postulates that structure can be described by a directed acyclic graph (DAG), which can be partially reconstructed by using the faithfulness principle. According to this principle, conditional independencies in the observational distribution correspond exactly to those encoded in the unknown DAG. However, such algorithms are challenged by the fact that, without even stronger assumptions, approximate failures of faithfulness are empirically undistinguishable from exact ones. This leaves an inferential gap which can lead to major implications to causal effect identification. We introduce a different perspective, not reliant on exact independencies. By considering the specialized problem of discovering whether unmeasured confounding can be ruled out for a particular cause-effect pair, we introduce a principle we call naturalness, in which the strength of observational associations bounds the strength of causal dependencies. The bounding function is unknown, with degrees of freedom not present in faithfulness. We model uncertainty about this function using a Bayesian framing which allows for the use of experiments from separate studies, grounding an adaptive causal discovery principle on historical experimental data. We demonstrate its usefulness in partial identification of causal effects, and discuss how the promising results can motivate new principles for general causal discovery.
Data pruning can substantially reduce training cost by removing examples that contribute little additional learning signal, but it has not been studied for DNA language-model pretraining. We introduce Sequence Pruning (SP), a dynamic pruning framework that removes genomic sequences as they become easy for the model to predict. SP uses per-sequence pretraining accuracy computed directly from the model's forward pass, requiring no separate scoring model or additional pass over the corpus. We evaluate SP across five pretraining settings spanning two corpora, three model families, and both MLM and CLM objectives, with downstream evaluation on five genomic benchmarks. SP reduces pretraining data processing by up to 96% while largely preserving downstream performance. At moderate savings, pruning improves performance by up to 1.2 MCC points over full-data pretraining; at the highest savings, pretraining uses only 4% of the baseline GPU hours with little performance loss. SP also outperforms the dynamic pruning methods InfoBatch and BPAS across most comparisons. Beyond accuracy, our framework supports alternative criteria such as perplexity and EL2N, although accuracy yields the strongest downstream performance at matched savings. These results show that genomic pretraining does not require repeatedly processing all available sequences and that dynamic sequence-level pruning can substantially improve its data and computational efficiency. Code is available at https://github.com/FaiasPromit/pruning-dna-sequence.
Most temporal graph learning methods reduce future prediction to discriminative inference over past interactions, focusing on edges or node labels for a largely persistent set of nodes. This view breaks down when new nodes and edges appear, since the future graph becomes a new object with changing size, composition, and structure. We address this gap by framing temporal graph prediction as an inverse topology problem. Instead of predicting edges directly, we first predict a multiscale topological descriptor of the future graph and then reconstruct a plausible future snapshot that realizes this descriptor under inductive constraints. This approach makes global structure and node churn explicit prediction targets and produces forecasted graphs that downstream tasks can evaluate without retraining. Across 14 temporal graph datasets, we evaluate our method, TOPOGED, on node, link, and graph property prediction and compare it against state-of-the-art temporal graph models. TOPOGED significantly improves node-level forecasting accuracy over the strongest baseline, with a macro average of 0.49 versus 0.05. It also outperforms baselines in 60% of graph-structure metric evaluations and substantially increases macro-average edge prediction metrics, from near-zero to 0.12. Our results show that topology-guided graph forecasting can predict inductive future snapshots whose structure supports multiple downstream evaluations.
LLM agents have made impressive progress in coding and mathematical reasoning, motivating growing interest in learning from long-horizon interaction. Memory supports this process, but plausible lessons can be redundant or reinforce inappropriate actions. Can a memory's source decision provide a practical test for whether to retain it? We localize consequential trajectory steps, extract candidate memories, and replay each source decision with and without its candidate under the same context. This reveals two properties: most candidates barely shift the probed distribution, and both successful and failed source actions can be reinforced or suppressed. These observations motivate Anchor, which tests shift magnitude and source-role alignment before retention. Admitted memories guide later trajectories, enabling experience-driven improvement under a frozen policy. On held-out ALFWorld and WebShop tasks, Anchor improves success rate and mean score by up to 19.9% and 17.2%, respectively, relative to the strongest baseline for each backbone. With Qwen3.5-27B, it retains 41% fewer ALFWorld cards than unfiltered retention and outperforms all four WebShop cost-comparison baselines while using fewer tokens.
Open-vocabulary 3D scene understanding aims to segment and retrieve arbitrary objects in 3D scenes using natural language queries. Most existing methods typically follow a splat-then-segment paradigm, where 3D Gaussian Splatting is first reconstructed and then augmented with labels or language features for subsequent segmentation. However, this pipeline often suffers from view inconsistency, ambiguous object boundaries, and weak object-level correspondence, especially in complex static and dynamic scenes. Building on our prior Segment-then-Splat framework, we present RefineSplat, a reconstruction-guided extension that retains its object-centric paradigm: it first tracks objects in 2D and then reconstructs each object with a dedicated set of 3D Gaussians, preserving explicit object-Gaussian correspondence for unified open-vocabulary segmentation and retrieval. To improve robustness, we further close the loop between 2D tracking and 3D reconstruction. Since off-the-shelf video trackers can produce unreliable masks that propagate errors into object reconstruction, we introduce a reconstruction-guided feedback mechanism that refines object tracking using rendered masks and multi-view consistency. Moreover, instead of assigning each object a single language embedding, we propose a viewpoint-aware embedding assignment strategy, where each object maintains multiple view-clustered embeddings to capture view-dependent semantic cues. We also introduce Blender-OVS, a synthetic benchmark of static and dynamic sequences with exact object meshes and point clouds, enabling direct evaluation of extracted geometry. Experiments on static and dynamic benchmarks show improved rendered-view segmentation and lower object-level Chamfer distances, with competitive scene reconstruction quality.
Wildland urban interface (WUI) fires increasingly threaten communities and infrastructure, motivating the need for rapid visual assessment. Recent research has focused primarily on detecting fire and assessing post-event damage. However, structure-level fire threat identification during active WUI fire events remains underexplored. This study evaluates Qwen3-VL-8B-Instruct for this active-fire decision using two separately scored tracks: 1,824 synthetic images from 491 designed WUI scenarios and 45 frames from a controlled-burn experiment conducted in collaboration with Columbus Fire & EMS Department. Under zero-shot prompting, the model returned valid target fields for 1,808 synthetic images and 39 controlled-burn frames. Accuracy among valid outputs was 94.0% for the synthetic track and 69.2% for the controlled-burn track, compared with a 16.0% random-selection baseline for the latter. On the synthetic track, the model generally identified the intended target across varied designed WUI scenarios when the relevant fire-spread cues were visually distinct. Controlled-burn performance was well above random selection, but errors showed difficulty interpreting spatial relations in complex scenes. The dual-track protocol provides scalable stress testing in designed scenarios and a controlled-burn reference for assessing model behavior under physical fire conditions. Future work can use the synthetic data for VLM post-training and then evaluate the adapted models on larger and more diverse real WUI fire datasets.