Papers

Generative AI makes options cheap and shifts the bottleneck of creative work from generating options to screening them. We argue that AI's value depends on whether its options arrive before or after the team's screen. In our model, early options compete for a place in the portfolio; late options arrive after that competition and are kept unscreened. In a preregistered experiment, we randomly assigned 148 three-person teams to five schedules of access to the same language model during a product-design task with timed search and selection stages. Moving the same 15 minutes of access from search to selection lowered expert-rated creativity by 0.62 standard deviations, although teams used the model at similar rates. Platform logs locate the difference at the screen. Early access enlarged the pool of ideas fivefold, and teams pruned it. Late-access teams reached the screen with the same small pool as teams without AI and then searched again, devoting 29% of their selection-stage queries to generating new options, compared with 11% for teams with access throughout. Late access added breadth, not quality: these teams submitted 73% more proposals than did teams without AI. Language-model judges detected the benefit of access but not the cost of late timing.

We investigate the influence of insoluble-surfactant elasticity and Boussinesq-Scriven surface viscosity on the parametric instability of an interface separating two viscous fluids. The two-fluid Floquet formulation of Kumar& Tuckerman (J. Fluid Mech., vol. 279, 1994, pp. 49-68) is extended to include surfactant transport, Marangoni stresses and surface viscosity. For the parameter range considered, both mechanisms increase the onset acceleration relative to the clean interface state, but through distinct responses: Marangoni elasticity produces a non-monotonic threshold, whereas surface viscosity yields a monotonic increase towards saturation. Analysis of the corresponding unforced gravity-capillary modes shows that both mechanisms modify modal damping and natural frequency. Wavenumber selection reduces the detuning from subharmonic resonance by approximately an order of magnitude relative to that evaluated at the clean interface critical wavenumber. Along the selected branches, variations in the onset acceleration closely follow those in modal damping, with residual detuning playing a secondary role. Despite their distinct interfacial stress mechanisms, both Marangoni elasticity and surface viscosity suppress tangential interfacial motion in the strong-rheology limit, approaching a common tangentially immobilized condition.

Recurrent algorithmic reasoners are trained with per-step hints on small graphs and run for many more steps on large ones. On breadth-first search in SALSA-CLRS, we find that a reasoner can reproduce every training hint and still fail on deep graphs: after some exact steps, it declares nodes reached before the search gets there. We call this a premature flood and trace two routes to it. In our GRU processor with a degree input, the flood is carried in the hidden state and set off when that input is in the tail of the training distribution. Setting the input higher at inference brings the flood sooner, and none of the six flooding models counts steps. Without the input, summing the neighbors' states instead of averaging them also breaks the processor. In the benchmark's own GIN and PGN, trained with its code, some seeds instead flood from a fixed step, on sparse and dense graphs alike. Clamping the degree input or adding 200 dense training graphs removes every failure of the GRU processor. Detaching the state between steps avoids the flood in that processor (0 of 23 seeds, against 6 under full backpropagation) but not with sums or a residual update, and two published remedies leave floods. None of the benchmark's test sets is both deep and dense, but its deepest one shows the second route: on its Delaunay graphs at 1,600 nodes, one GIN seed floods from step 18. We therefore recommend auditing the range of every input and aggregate, which catches the first route, and testing on deep graphs, which catches both.

We prove Kahane's conjecture: if a continuous unimodular function $F$ on the circle satisfies $\|F^n\|_{A(\mathbb{T})}=o(\log|n|)$ as $|n|\to\infty$, $n\in\mathbb{Z}$, then $F(x)=e^{i(mx+b)}$ for some $m\in\mathbb{Z}$ and $b\in\mathbb{R}$. The main step is a logarithmic lower bound for averages: for every nonconstant real-valued $\varphi\in A(\mathbb{T})$, $\frac{1}{T}\int_a^{a+T}\|e^{it\varphi}\|_{A(\mathbb{T})}dt \ge c_\varphi\log T \quad(a\in\mathbb{R},\ T\ge T_\varphi)$, where $c_\varphi>0$ and $T_\varphi>1$ are independent of $a$. The proof uses exact sampling on finite subgroups of the circle, uniformly bounded localized modulations, and the weighted non-harmonic Littlewood inequality of Jaming–Kellay–Saba. A suitable cutoff and a small linear perturbation of the phase ensure that a fixed positive proportion of the coefficients at the beginning of the frequency ordering have modulus one. Passing to integer powers and removing the degree of the circle map yield Kahane's conjecture without any additional smoothness assumptions. We also give equivalent phase, convolution, and operator formulations, and derive consequences for homomorphisms of weighted Wiener algebras and for higher-dimensional tori.

We present a generator of Thai klon pad verse, one stanza from a short topic, and evaluate it against human verse on held-out topics. The generator is a masked language model initialised from WangchanBERTa over a fixed canvas with two layers: a sound layer of pronounced syllables, where rhyme, tone and syllable count are enforced as hard masks, and a word layer of whole words, so that no word is cut into written pieces. Decoding commits one cell at a time and samples at temperature 0.4; one sample is drawn per topic, with no selection, reranking or repair. On 78 held-out topics never used in development, judged by three independent runs of a frozen LLM judge in the same pools as the human stanzas from which the topics were written, 70.5% of the generated stanzas pass form, lexical cleanness of all four lines and topic alignment together, against 76.9% for the human stanzas (difference -6.4 points, 95% CI [-19.2, +6.4]). Cleanness is level with human (85.9% vs. 87.2%), and the stanzas pass form and cleanness together slightly more often than the human ones (+3.8 points, [-6.4, +14.1]); the difference lies in alignment (84.6% vs. 91.0%), on topics that were written from the human stanzas. The low temperature matters: at higher temperatures the same model passes clearly fewer of the same topics, because lexical errors lie in the model's lower-probability choices; sharpening the sampling distribution removes most of them, at a cost in vocabulary (7.5% rare words against 16.5% in the human stanzas) and diversity. The pre-registered target, all-three within 5 points of human, is missed by 1.4 points. KlonSuphap-LM, the only public klon pad generator we know of, passes none of the 78 held-out topics when judged in separate pools together with the method's and the human stanzas (form 19.2%, cleanness 3.8%).

Let $D$ be a finite simple minimally strongly $k$-connected digraph, where $k\ge 2$, and let $n=|V(D)|$. Mader proved in 1985 that $|A(D)|\le 2k(n-k)$ for all sufficiently large $n$, with equality if and only if $D\cong \overleftrightarrow{K}_{k,n-k}$ (JCTB 1985). We prove that this bound holds whenever $n\ge 3k-1$ and that the order threshold $3k-1$ is best possible. Moreover, if $n\ge 3k$, then equality holds if and only if $D\cong \overleftrightarrow{K}_{k,n-k}$. The threshold $3k$ is also best possible. These results resolve the conjecture of Mader in (JCTB 1985).

This paper is concerned with the inverse scattering problem of determining a real-valued potential for the three-dimensional Schrödinger equation from multi-wavenumber data. Under suitable a priori bounds, we establish local increasing stability from near-field Dirichlet data on a fixed sphere for polynomially decaying potentials, and global increasing stability from full far-field amplitudes for exponentially decaying potentials. The stability estimates consist of a noise term and a remainder term, with the latter decreasing as the upper measured wavenumber increases. We also obtain global uniqueness from exact near-field data under weaker regularity assumptions. The analysis combines analytic continuation, weighted resolvent estimates, and half-ray identities whose differences at chord endpoints cancel the unknown exterior contribution.

Embodied agents built on foundation models rely on an execution harness to organize observations, guide actions, and incorporate execution feedback into subsequent decisions. Failures can arise not only from limited model capabilities, but also from how the harness presents evidence and guides its use. These model-external mechanisms provide an adaptation target without updating model parameters, but task-specific interaction experience can be costly to collect. Task-level outcomes alone do not reveal whether relevant evidence was missing from the model input or was available but inadequately used. Moreover, these two aspects are coupled: effective action guidance may require evidence that the current context omits, while exposing that evidence alone does not ensure an appropriate response. We introduce EnactOpt, which uses limited execution feedback to jointly evolve reusable textual skills and context-construction code while keeping the underlying models and runtime fixed. It diagnoses adjacent decisions by relating action feedback to the actual model input and subsequent action. Structured episode summaries guide joint revisions through prompt-level edit operators, and candidate harnesses undergo compatibility checks and validation-based retention. Across VLABench, ALFWorld, and Embodied Agent Interface (EAI), EnactOpt improves task success by approximately 10-27 percentage points over the initial harnesses and outperforms the evaluated GEPA, TextGrad, and Trace2Skill baselines. These gains extend to task conditions not used for harness evolution or candidate selection. The results support a complementary route to embodied adaptation: turning limited execution experience into reusable behavioral improvements by jointly refining how evidence is presented and acted upon.

Large Vision-Language Models (LVLMs) have excelled in joint visual and language understanding, particularly in image captioning. However, LVLMs remain prone to object hallucination, describing non-existent entities in the given image. Previous remedies, including dataset-enhanced supervised training and reinforcement learning from human feedback, show promise but rely heavily on costly human involvement, limiting scalability. To address this, we propose a reinforcement learning framework to mitigate object hallucination in LVLM-based image captioning, driven entirely by natural language processing (NLP) metrics. While optimizing reinforcement learning solely on NLP metrics offers scalability, it also introduces pitfalls: reward hacking, sparse feedback, and costly training. Our framework sidesteps these by introducing an Object Consistency Score (OCS) capturing both correctness and coverage of objects, incorporating KL-regularization for denser token-level guidance, and designing a lightweight PPO variant that fine-tunes adapters atop a frozen LVLM backbone, achieving efficiency and stability. Extensive experiments on the InstructBLIP baseline demonstrate the effectiveness and generalizability of our framework, achieving a substantial hallucination reduction of up to 41% on the COCO dataset and 29% on the Visual Genome. Furthermore, the framework can enhance caption quality by incorporating complementary evaluation metrics such as BERTScore and METEOR. These results reveal a broader insight in multimodal learning: LVLMs can effectively self-correct hallucinations using metric feedback, motivating deeper exploration into scalable reinforcement frameworks that significantly reduce human effort.

We provide a unified algebraic classification of approximate controllability, approximate null controllability, exact null controllability, and exact controllability for finite-dimensional linear stochastic control systems with constant coefficients and multidimensional Brownian noise. The controls may enter both the drift and the diffusion terms, and the joint diffusion-control matrix may have arbitrary rank. A finite decreasing sequence of subspaces defines a common geometric obstruction. Its vanishing characterizes the two approximate properties and exact null controllability. Exact controllability to arbitrary random targets additionally requires the joint diffusion-control matrix to be surjective. The codimension of the obstruction is given by a rank difference of explicit polynomial block matrices, yielding finite tests in the original coefficients. When the joint diffusion-control matrix is surjective, we obtain a finite Kalman-type full-rank condition for exact controllability. If the obstruction vanishes, a linear state feedback with a deterministic time-dependent gain drives every initial state to zero at any prescribed positive time with finite energy and polynomial small-time cost. An equivalent orthogonal elimination algorithm computes the rank differences with polynomial storage and cost, avoiding exponentially large block matrices. A two-noise example relates the input classification to terminal errors and control energy in penalized stochastic linear-quadratic control problems.

Optogenetics enables methanol-free protein production in the yeast $\textit{Komagataella phaffii}$ using light as a control input, facilitated by the blue-light-responsive transcription factor EL222. This opens the door to more programmable protein production in the food and biopharmaceutical industries. However, excessive illumination can inhibit growth and protein accumulation in some of these engineered cell factories. Previous hybrid modelling used black-box regression across sparsely sampled conditions to infer light-dependent kinetic parameters, without explicitly representing intracellular regulation. Here, we develop a multiscale mechanistic model linking EL222 activation, transcription, and translation to protein, biomass, and substrate dynamics. A latent light-associated stress state represents photoinhibitory effects on growth, transcription, and translation. An ensemble-based parameter-estimation strategy combines repeated global-local optimisation with screening for parameter sets supporting locally stable intracellular steady states under specified light conditions. Using available batch data for yeast-enhanced green fluorescent protein production in a $\textit{K. phaffii}$ strain carrying eight EL222 gene copies, the ensemble-mean predictions captured the main experimental trends in protein accumulation and growth across six constant light intensities. Parameter variability was propagated to time-course predictions, providing an empirical assessment of parameter-related uncertainty. Wider prediction bands coincided with less complete experimental coverage of the growth phases. Overall, the framework provides a mechanistic basis for future model-based optimisation and control of optogenetic bioprocesses under phototoxicity.

Let $(A,\mathfrak{m})$ be a Noetherian local ring and let $I\subseteq \mathfrak{m}$ be an ideal in $A$. Let $\ell(I)$ be the analytic spread of $I$. Let $G_I(A) = \bigoplus_{n \geq 0}I^n/I^{n+1}$ be the associated graded ring of $I$ and let its depth be $t$. We prove $\text{depth} \ A/I^n \geq t - \ell(I)$ for all $n\geq 1$.

We study uniqueness for parabolic motions in the Newtonian $N$-body problem with a prescribed minimal normalized central configuration at infinity. A McGehee compactification identifies their tails with the strong stable manifold of the corresponding outgoing zero-energy equilibrium. The configuration projection of this manifold is locally a smooth graph, yielding a smooth initial-velocity field and uniqueness among all parabolic motions confined to a sufficiently narrow truncated cone, without any minimizing assumption. For parabolic geodesic rays, a renormalized action excess and a quantitative no-exit argument remove the confinement hypothesis when the initial configuration lies sufficiently far inside a smaller cone. We then prove that the set of collision-free initial configurations admitting more than one such ray has Lebesgue measure zero. The global argument propagates the local stable manifold backward, discards critical values of the configuration projections by Sard's theorem, and separates distinct regular branches through an explicit differential formula for their renormalized excess. No nondegeneracy assumption on the prescribed minimal central configuration is required. The proofs combine invariant-manifold theory with direct action comparisons and do not use weak KAM theory, Busemann functions, or the differentiability theory of Hamilton–Jacobi solutions.

WavePhaseNet introduced frequency-resolved semantic representations, QuantumPhaseNet introduced gauge-covariant phase connections, and ModularPhaseNet made a finite cyclic phase channel and its cycle holonomy computable in an ordinary Transformer. These constructions are complementary, yet they do not by themselves separate an observed collection of pairwise relations into a globally consistent direction, circulation around local compositions, and circulation that persists around a larger unfilled cycle. HodgePhaseNet addresses this separation by representing semantic transitions as an antisymmetric real-valued 1-cochain on a finite oriented simplicial complex. With vertex-edge incidence matrix $B_1$ and edge-face incidence matrix $B_2$, every such field has a unique orthogonal decomposition $\Omega = B_1^T\Phi + B_2\Psi + H$, where $B_1H=0$ and $B_2^TH=0$. The three component fields are unique even when their potentials are not. We give explicit projectors, a sparse implementation, integrability and gauge results, a verified numerical example, and a redesigned attention interface. The attention bias is defined on allowed ordered query-key pairs, remains consistent under arbitrary choices of edge orientation, and preserves a decoder's causal information boundary by constructing its complex from the visible prefix. Gauge invariance of the attention output is conditional on using invariant or appropriately compensated features. A simple difference of vertex embeddings has no coexact or harmonic component; a genuine relation encoder is therefore indispensable. Local circulation is not automatically contradiction, and harmonic flow is not automatically missing knowledge. These interpretations and any reduction in factual error require controlled empirical evaluation. This paper presents a mathematical and architectural proposal, not external natural-language performance results.

With the rapid evolution of large language models (LLMs), there is an increasing demand for effective personalization techniques that can adapt model behavior to individual user preferences. Personalized LLMs must infer a user's query-dependent preferences from limited, heterogeneous, and evolving interaction histories. Existing approaches rely primarily on in-context personalization, parametric adaptation, or personalized reward modeling, but respectively incur recurring costs in context management, model optimization and deployment, or multi-candidate generation and scoring. These limitations become particularly restrictive when user observations are sparse and the preference evidence relevant to each query varies across contexts and over time. To address these challenges, we propose \modelname, a fine-grained and instance-tailored steering framework that dynamically generates sample-level interference vectors from user data and injects them into the model's forward pass for personalized adaptation. Our approach introduces: match contrast that produces user-relevant signals, fine-grained steering component that captures nuanced signals by hooking activations from attention and MLP layers, and query-conditioned aggregation module that synthesizes these signals into contextually relevant enhancements. Extensive experiments across diverse scenarios, including short-to-long text generation, and web function calling, validate the effectiveness and compatibility of \modelname. Results show that our method consistently brings significant performance gains, and show extreme superiority in fast-shifting and data-sparse environments. Additionally, \modelname rids us of training cycles and incurs light overhead of storage or inference latency. These make \modelname a strong and real-world deployable method. Implementation is available \footnote{https://anonymous.4open.science/r/FleetPilot}.

Innovation changes welfare by improving technology and reallocating inputs across producers with different markups. When producers choose markups optimally, an innovator's own output and price adjustments have no private value at the margin, so reallocation affects its profit only through others' responses. For an innovator small in its markets, the change in allocative efficiency is of the same order as its cost saving but does not reach its profit. A process innovator is paid its cost saving, forgoing its own and downstream markups on added output. For U.S. public firms in 2015, these forgone markups average 26 percent of that saving. For innovators accounting for 56 percent of sales, allocative efficiency rises or falls by more than a tenth of their private return. With estimated spillovers, patenting firms' R&D has a social return 6.24 times its private return.

Large Language Models (LLMs) can generate functionally correct code while introducing exploitable security weaknesses, particularly when developers specify functionality without explicit security requirements. Recent state-of-the-art approaches mitigate this problem using external security knowledge, but largely rely on flat similarity-based retrieval over independent artifacts, making it difficult to infer the relevant vulnerability space and retrieve targeted knowledge when direct matches are unavailable. We propose SEEK, an agentic Graph RAG framework for secure code generation that constructs a security knowledge graph using the Common Weakness Enumeration (CWE) taxonomy as its structural backbone and injects it with curated real-world CVE vulnerability-fix instances. At inference time, a Planner Agent predicts potential CWE weaknesses associated with the developer query, which guide Graph Scoping to restrict retrieval to relevant regions of the CWE taxonomy before deterministic multi-faceted candidate ranking. A Selection Agent subsequently validates the predicted weaknesses and retrieved candidates prior to code generation. Across five open-source LLMs, we evaluate SEEK in an out-of-distribution setting on CodeGuard+ and CyberSecEval, spanning C, C++, and Python. SEEK achieves the highest SecurePass@1 of 77.45% on CodeGuard+ and the highest aggregate ICD pass rate of 79.45% on CyberSecEval, surpassing prior approaches. Moreover, SEEK achieves 85.2% SecurePass@1 across 13 unseen CWE types with no directly linked vulnerability-fix instances, demonstrating strong generalizability, while maintaining acceptable inference latency costs.

Soft matter systems often display shape transitions, mode competition, and threshold-like dynamical responses driven by the coupling between mechanics, geometry, and biochemical or active fields. Here we present a minimal theoretical framework for describing such morphodynamic selection using a small number of physically interpretable order parameters. The model combines elastic relaxation, active driving, and nonlinear feedback, yielding distinct regimes corresponding to weak deformation, single-mode selection, and coupled shape transformation. We derive the governing equations, identify the relevant dimensionless control parameters, and show how the phase behavior can be organized in terms of threshold competition and effective coupling strengths. The framework provides a compact route for connecting analytical theory, numerical simulations, and experimentally observable shape dynamics in soft and biological matter.

Long-horizon agents must complete workflows in which later activities depend on earlier outputs and changes to shared state. A key challenge for evaluation is to expand task requirements and their dependencies while keeping tasks executable and outcomes verifiable. We introduce SkillHorizon, a benchmark that constructs long-horizon tasks through skill composition in seven business environments. Its generators use domain-specific rules to turn reusable skill procedures into concrete business activities. These activities are connected through input-output dependencies, shared resources, and state updates, allowing workload and coordination demands to grow together. Successful reference executions validate task feasibility and measure interaction workload, while programmatic evaluators assess local completion and whole-task success. We evaluate five language models on 1,050 tasks with reference interaction workloads ranging from 64K to 1M tokens. Across these scales, mean local-goal completion falls from 90.4% to 46.8%, as completed work fails to keep pace with growing requirements. At larger workloads, agents more often submit unsuccessful final answers with substantial work unfinished and tool-call allowance remaining. Trace analysis also identifies cases where scripts execute successfully but their outputs fail downstream business checks. SkillHorizon provides a configurable testbed for studying sustained progress, dependency coordination, and complete task delivery in growing workflows. Our code is available at https://github.com/KnowComp/SkillHorizon.

We formulate offline tuning of Model Predictive Path Integral (MPPI) controllers as a chance-constrained black-box optimization problem and develop an update-aware framework based on Matrix Adaptation Evolution Strategy (MA-ES). We use statistical information from noisy controller evaluations to determine when the evolutionary update is sufficiently reliable. When the recombination direction is certified, a derivative-free nonmonotone line search (NMLS) is used to refine the candidate solution, and a separate validation stage assesses the final frozen controller. We establish a worst-case arithmetic-complexity bound for both online MPPI control and offline tuning. Numerical experiments on TurtleBot 4, Jackal-like, and Ackermann models show that the proposed NMLS refinement is competitive with standard MA-ES, transfers across the three vehicle models without model-specific retuning, and improves several cost and validation measures. Both MA-ES variants also compare favorably with the tested randomized derivative-free line search (VRBBO) configuration. The results support the practical applicability and transferability of the proposed framework, while not claiming universal superiority or hardware-level safety.

Cross-material magnetic core loss prediction requires adapting to new material responses with limited measurements. Global feature alignment overlooks differences in source and target excitation coverage. This motivates selecting comparable samples and controlling each source measurement's alignment contribution. This paper proposes T-PCOT, a Transformer transfer method based on condition-supported partial optimal transport. Frequency, flux excursion, temperature, and waveform descriptors define a sparse graph of comparable operating conditions. A capacity-constrained transport plan distributes alignment across source measurements and leaves costly target mass unmatched. The plan is constructed from measured inputs before training and guides local waveform-feature consistency. An importance-weighted sampling scheme preserves both uniform source regression and the transport objective during minibatch optimization. Experiments on three material pairs and three target label budgets show that T-PCOT reduces mean relative error (MRE) and 95th-percentile error (P95) relative to Transformer-based multiple-kernel maximum mean discrepancy (T-MKMMD) in all nine comparisons. Macro MRE decreases from 12.60% to 9.65%, and macro P95 from 30.94% to 25.06%, corresponding to relative reductions of 23.41% and 19.01%. Controlled comparisons and ablations examine the roles of input processing, condition correspondence, and source capacity. The results demonstrate the predictive advantage of organizing cross-material alignment through operating-condition support over the compared global alignment method. The proposed method achieves state-of-the-art predictive performance in magnetic core loss transfer learning.

Sparse-view Cone-Beam Computed Tomography (CBCT) reconstruction is essential for reducing radiation dose and scan effort, but remains underconstrained due to limited projection information. Recent 3D Gaussian Splatting (3DGS)-based CBCT reconstruction methods rely on gradient-driven densification, which becomes unstable under sparse acquisition and often leads to over-smoothed structures. We propose AD-GS, an Anatomical Density-guided 3D Gaussian Splatting framework that formulates Gaussian densification as an attenuation-aware Markov Chain Monte Carlo (MCMC) sampling process. By interpreting Gaussians as samples from an underlying anatomical attenuation distribution, AD-GS adaptively concentrates representation capacity in high-attenuation regions, enabling sharper structural boundaries under sparse projections. To further constrain reconstruction, we incorporate a semantic regularization based on zero-shot features from Segment Anything in Medical Images, enforcing multi-view consistency through semantic priors. Experiments on the ToothFairy3 dental CBCT dataset demonstrate that AD-GS achieves superior structural fidelity and boundary preservation compared with existing 3DGS-based approaches under sparse-view settings within the dental and maxillofacial setting. Code is available at https://github.com/yulinjiaxue/AD-GS-BMVC2026.

We study a non-convex, nonlinear integer mathematical program with a linear constraint. This problem originates in finance and consists of selling a large predetermined quantity of a financial asset over a fixed time horizon, while maximizing the total proceeds. The selling price is modeled using a deterministic penalty function capturing the market's adverse reaction to the selling flow. From a complexity perspective, the problem is in NP. We first solve the problem exactly using dynamic programming and propose several heuristics that yield tight numerical lower bounds. We then reformulate the problem into an equivalent factorable form and derive a tight convex relaxation using McCormick inequalities. This relaxation is embedded within an adapted Branch & Bound algorithm to find the global optimum in finite time. Numerical experiments show that our methods outperform several widely used generic solvers (LocalSolver, Scip, Baron, Couenne, and Bonmin) for selected instances, but do not yield tight upper bounds, beyond small instances.

We present HiPAD, a single fusion network submitted to all six sub-tracks of the MPDD-AVG 2026 depression-detection challenge, covering the Elder and Young cohorts and three modality sets (gait, audio-visual, and both), each combined with personality descriptions. The network encodes each available modality with a shared BiLSTM, zero-pads absent modalities, and applies cross-modal attention only when at least two of gait, audio and video are present. Personality, which dominates our results, is kept out of the attention and fused by concatenation. Because the architecture is identical in every sub-track, differences between them reflect the data rather than the model, and we use this to study three questions. Our main finding is that the personality description, roughly 150 tokens about who the subject is, carries more depression signal than audio, video, and gait combined: gait plus personality alone (TLS 0.647) beats every configuration that includes audio and video. First, a four-level ablation of the personality description (L0-L3) quantifies this: removing every personality-derived feature, including the Big Five scores that earlier MPDD ablations keep, lowers the Track-Level Score by 0.499 to 0.507 and leaves the model at chance. Second, domain-adapting the personality encoder with LoRA (7B) or QLoRA (72B) on PHQ-8/9 data raises binary Cohen's $\kappa$ by 0.13 but lowers the overall score by distorting the PHQ-9 regression; a frozen RoBERTa-large encoder does better at this sample size. Third, at fixed architecture, handcrafted clinical features (eGeMAPS, OpenFace action units) outperform generic deep features (wav2vec2, DenseNet) on the audio-visual tracks. The unified system reaches a cross-track G-Score of 0.482; with under 100 subjects per cohort, the limiting factors are the size of the representation and the calibration of the outputs, not the choice of architecture.

Can an autonomous AI agent reverse engineer a binary as well as a skilled human analyst? Do they adopt the same strategy, and can agents scale to real-world programs? We compare human and LLM-based software reverse engineering (SRE): nine models tackle two compact static-analysis challenges previously solved by human analysts, scored with a common understanding rubric, and we record their complete interactions with IDA Pro. This lets us compare not only what humans and agents recover, but also how they investigate a binary. The best agents (GPT-5.5 Pro, Claude Opus 4.8, and Kimi K3) achieve statistically human-competitive understanding on both tasks, while weaker models remain far behind. Ablations show that the gap depends on code representation and workflow demands, not on model capability alone: open-weight models improve substantially with richer code representations, whereas frontier models show no significant change. Similar success rates hide different workflows. Humans repeatedly revisit functions and follow local program structure; strong agents jump across the binary, retrieve information broadly, and rarely return to inspected code, a behavior we call greedy fetching. This strategy is not fixed: given semantically richer code, agents fetch less without losing understanding. Because greedy fetching aggregates large amounts of code in the agent's context, we measure context-window saturation across 4,000 real-world binaries: large windows accommodate many programs, but larger applications and lower-level representations quickly exhaust them. Autonomous reverse engineering is becoming competitive, with a large speed advantage, but scaling it will require better navigation, selective retrieval, and persistent analysis state, not merely larger context windows.

Formulaic alpha mining seeks to discover executable expressions that capture predictive market signals in a vast, noisy search space. Recent LLM-based approaches use prompting, iterative search, or seed-factor evolution, leaving direct factor construction through policy learning less explored. To address this, we introduce AlphaRLM, a reinforcement-learning framework that trains LLMs to construct executable symbolic factors from financial literature without input factor expressions. AlphaRLM incorporates Literature-Guided Operator Assembly (LEGO), which conditions factor construction on literature-derived ideas and executable operators, providing semantic priors for exploration. It further aligns the generation policy with market feedback through executable-validity and grouped-return-separation rewards that capture non-monotonic structures across factor quantiles. Extensive experiments on CSI 300, CSI 1000, and S&P 500 show that AlphaRLM outperforms strong factor-mining and LLM-based baselines in annualized excess return. Notably, the 8B model achieves strong cross-market performance, highlighting the promise of literature-conditioned, market-aligned factor generation.

Activation steering has recently been used as a memory for skills: past experiences of an agent are stored as steering vectors and retrieved later. Facts and preferences about the user are assumed to need text instead. We test that assumption with a training-free activation memory beside a frozen LLM. The change an experience causes in late-layer states is stored as a shift, filed under a whitened key for the conversation's gist, and added back in a later session only when the gist matches. We argue that such a shift is a fixed average of an effect that attention, in context, recomputes for every question, so only question-independent content can survive. In a pre-registered study on Qwen3.5-2B and 9B, with judges from four model families, the memory is selective (about 98% of unrelated replies are unchanged), but facts are stored as the word to say, in the last layers only, and are almost never used (at most 15% clean use, against 96-98% in context). Leanings transfer only in a narrow strength window. Most strikingly, the reference subtracted when writing a preference decides whether its concept or its direction survives: unless the opposite is subtracted, a stored dislike is recalled as a like.

Recent video generative models have shown strong capability in synthesizing high-quality and temporally coherent videos. Yet, adapting them for controllable portrait animation remains nontrivial, as existing methods typically rely on architectural changes or extra conditioning modules, increasing complexity and reducing compatibility with pretrained models. We present CatPortrait, a simple and effective framework that enables controllable portrait animation via 3D face guidance as in-context conditioning. CatPortrait preserves the original video generation backbone and injects structured 3D facial guidance into the input space through spatial concatenation, allowing pretrained image-to-video models to animate portraits without sacrificing their pretrained priors. CatPortrait supports both video-driven and audio-driven portrait animation within a unified framework. Extensive experiments demonstrate that CatPortrait achieves competitive or superior results across both settings.

The Model Context Protocol provides a shared interaction substrate for the agentic web, allowing heterogeneous services to expose capabilities through a standardized interface. Learning to use this ecosystem entails a hierarchy of capabilities: tool-format skills, server-level transfer, and cross-server composition. We present MCPDisCo, an ecosystem-scale reinforcement learning framework for MCP-native agents. Using a Thought-Markov Decision Process (Thought-MDP), the framework explicitly represents internal reasoning alongside routing, execution, and termination and trains on real-server, long-horizon interactions spanning 70 MCP servers and 527 tools across six domains. We study the reasoning–tool seesaw, a joint-learning failure mode in which improvement in one objective can accompany deterioration in the other. To mitigate potential interference, MCPDisCo combines token-masked reward attribution with bidirectional gradient surgery: the former separates direct supervision for reasoning and action spans, while the latter removes conflicting components from both objective gradients in shared parameters. MCPDisCo achieves higher success with fewer chain-of-thought tokens. On the out-of-distribution MCP-Atlas and the GAIA benchmark, MCPDisCo maintains monotone gains, confirming that the learned behavior generalizes beyond the training server distribution and broader agentic tasks. Model weight and implementation are available.

Rhythm games ask the player to act in time with a recording, following a chart: a sequence of actions, each with a time and, depending on the game, a lane, type, or playfield position. Charts are hand-written per recording and difficulty. Automatic generation from audio has been studied with pipeline and end-to-end approaches, and recent work reuses pre-trained audio encoders, but never compares them with one trained from scratch. None of these targets automatic music transcription (AMT), yet most chart times are note onsets, which AMT models locate. We freeze a pre-trained audio encoder, an AMT model in particular, and train only a linear projection and a Transformer decoder on charts. With MuScriptor, this fits human charts more closely than encoders trained from scratch and places more accurate note types than MERT, a music encoder pre-trained without transcription labels. Joint training on three osu! modes raises event F1 on the mode with the fewest charts from 0.492 to 0.506, while charts from Clone Hero, similar in layout, lower it to 0.482. Transfer between games thus appears to depend on charting style and the music covered.