Papers

Traffic microsimulators rely on hand-crafted behavior models that reproduce aggregate flow but miss the heterogeneous interactions between vehicles at signalized intersections. Learned trajectory predictors capture richer interactions but are short-horizon and tend to be unstable when run in closed loop. We present Enactor, an actor-centric generative model for closed-loop intersection microsimulation. The model focuses on vehicles; pedestrians are included as context that can influence vehicle decisions but not predicted. Dynamic actors and lane polylines are encoded in polar coordinates referenced to the intersection center. A transformer with separate spatial and temporal attention blocks predicts a distribution over each actor's next-step motion ($s$, $\alpha$). Training uses a closed-loop curriculum so the model is exposed to its own predictions. We evaluate Enactor in two regimes. In a 4000-second simulation-in-the-loop test at two intersection geometries, Enactor controls every dynamic vehicle against a continuously refreshing actor set rather than the fixed cohort that learned trajectory predictors are usually evaluated against. It recovers the SUMO data generator's speed and travel-time distributions with KL divergence over an order of magnitude lower than a recent transformer baseline on travel time, and substantially lower on speed (roughly $5\times$ lower at Site 1), and reduces red-light violations relative to the same baseline by more than an order of magnitude. An ablation isolates the leader rear-bumper feature as the change with the largest effect on intersection-aware safety metrics. We also evaluate on real-world field data and apply the same architecture to naturalistic vehicle trajectories from a fish-eye camera at a signalized intersection and evaluate it on multi-horizon predictive tasks. Enactor outperforms a constant-velocity baseline at every horizon evaluated.

As autonomous aircraft are introduced at scale and traffic density increases, centralized management becomes insufficient to coordinate the large numbers of crewed and uncrewed aircraft. Dedicated Advanced Air Mobility (AAM) corridors have therefore been proposed for organizing high-density autonomous traffic flows. The desire to scalably provide autonomous aircraft flexibility in trajectory planning motivates the development of decentralized approaches to traffic management in AAM corridors. In this work, we extend a multi-agent reinforcement learning (MARL) approach to address the challenge of decentralized traffic flow management in air corridor networks. We test policies trained in a single-corridor setting on increasingly complex multi-corridor networks with combinations of merges and splits in a zero-shot manner. Experimental results demonstrate that learned behaviors transfer well to scenarios with varying traffic density, network geometry, and heterogeneous vehicle performance, without needing centralized coordination or model retraining. We evaluate system-level performance in terms of conformance to corridor boundaries, completion rates, average speeds, distance traveled, and maintenance of inter-aircraft separation. We find that although our policies require only locally coordinated entry, traversal, and exit behaviors, they collectively produce desirable traffic flows through the corridor network.

Multimodal agents repeatedly re-examine the same video frames, UI screenshots, and rendered artifacts as their context window slides and reasoning iterates, yet every look-back re-encodes from scratch, because prefix caches serve reuse only at a fixed leading position. We show this recompute is avoidable, and identify exactly what naive KV reuse loses: the cross-chunk conditioning a chunk absorbs from its neighbours. This loss is asymmetric. The direct readout of a cached chunk is recovered exactly and for free by the standard state-merge. What remains is a diffuse, low-rank residue concentrated in deep layers, invisible to single-hop retrieval but precisely what multi-hop reasoning binds on. Blind reuse therefore leaves single-hop recall intact while halving multi-hop accuracy; this is the failure mode prior position-independent caches, designed for single-context or single-image reuse, do not address. We repair it with a small, training-free low-rank conditioning patch stored alongside each position-free chunk. Reuse reduces to one operator across MLA, GQA, and MHA: exact RoPE re-rotation to any target position, plus the patch that restores cross-chunk binding. This makes three window operations cheap: reorder (one patch serves every ordering of a cached set), sliding-window survival (surviving chunks relocate via rotation only, zero re-encode), and recall (an evicted chunk is rehydrated by its patch, never re-encoded). A rank-m patch recovers full task accuracy on cross-chunk-binding benchmarks, MM-NIAH across two attention families and two-page doc-QA, at a fraction of the KV footprint, and reconstructs re-prefill KV to within bf16 rounding in a production SGLang kernel across six backbones. The conditioning signal is strongest in redundant vision and video streams, making our solution most impactful where multimodal agents spend their recompute budget.

Hyperparameter tuning almost always means search: fit the model at every value on a grid, score each by cross-validation, and keep the winner. For spline regression that search is unnecessary. The optimal resolution can be solved for in closed form, to the accuracy an exhaustive search reaches, at a fraction of the compute. Three ingredients make this possible: classical approximation theory pins the squared bias to a known power of the resolution G, exactly the Kolmogorov n-width of the smoothness class; the basis dimension is an explicit polynomial in G; and leave-one-out error follows from a single fit via the PRESS identity. Balancing the two known curves gives the minimizer analytically. We extend this calculus to many coordinates by replacing ambient input dimension with interaction order, the number of active low-order components in an ANOVA decomposition, yielding a scaling law in which the optimal resolution and error are power functions of the effective density (sample size per active component), with input dimension absent from the exponent. The law becomes an algorithm. KORE (Kolmogorov-optimal Order-aware Resolution Estimation) fits two pilot resolutions, solves a leverage-calibrated 2x2 system for the bias and noise scales, and evaluates the closed-form plug-in resolution with a tiny leave-one-out certificate: about a dozen fits instead of a full grid sweep, with a consistency guarantee as the sample grows. Across additive and sparse pairwise targets up to 80 input dimensions, KORE matches exhaustive 3-fold cross-validation and the full classical ladder (GCV, Mallows' Cp, AIC, BIC) while fitting roughly 8x fewer models; on 36 real tabular datasets it ranks first among 21 methods in accuracy per unit of compute, ahead of tuned boosters and kernel machines. When complexity lives in low interaction order, solving for the resolution beats searching for it.

Large language models (LLMs) are increasingly being integrated into first-principles computational workflows, yet their ability to configure scientific calculations remains poorly understood. Here, we introduce INCARBench, a benchmark for evaluating LLMs on input configuration for the Vienna Ab initio Simulation Package (VASP) through both configuration generation and repair tasks. Evaluating 19 model configurations reveals substantial capability differences among current frontier models. While several models achieve high semantic and policy accuracy, task-critical correctness remains substantially lower, demonstrating that parameter-level correctness does not necessarily imply scientifically valid configurations. Failure analysis shows that errors concentrate in physically coupled settings involving DFT+$U$, magnetism, and correlated materials, where multiple constraints must be satisfied simultaneously. Repair evaluation further reveals that correcting incorrect settings and preserving already-valid configurations are distinct capabilities, with configuration preservation remaining a major challenge. These findings establish scientific configuration as a measurable capability of large language models and provide a foundation for developing more reliable AI systems for computational materials science.

Masked diffusion language models decode by iteratively unmasking tokens, where the unmasking order defines an "order of thought" that strongly influences generation quality yet is typically chosen heuristically. We derive a tractable upper bound on the sequential decoding mismatch, measured by the Kullback-Leibler divergence and expressed in terms of the model's pathwise log-likelihood, with tightness under sufficient model expressivity. This bound induces a dense self-aware reward over ordered trajectories, casting order selection as a principled policy optimization problem with a frozen denoiser. We instantiate this idea as Self-Aware Scheduling (SAS), which learns a lightweight order policy using Group Relative Policy Optimization and applies seamlessly to both any-order and semi-autoregressive decoding. On Sudoku with 1B MDM, SAS improves puzzle accuracy from 82.0% (best heuristic schedule) to 91.8%, and reaches 97.5% with second-stage fine-tuning along learned trajectories. On mathematical reasoning with LLaDA-8B, SAS improves pass@1 on GSM8K from 64% to 76% and on MBPP from 39.5% to 41%, consistently matching or exceeding heuristic schedules across generation lengths and block sizes. Project page: https://jimmyxu123.github.io/SAS

Transformer-based models underpin modern natural language processing but incur rapidly growing computational and energy costs. As training scales in both model size and parallelism, accurately predicting energy consumption has become critical for sustainable and cost-aware system design. We present a framework for modeling the energy consumption of Transformer training on multiple GPUs. Using controlled architectural sweeps of BERT models, we relate measured energy to lightweight proxies for compute, memory traffic, and hardware efficiency. Inspired by roofline models, our approach incorporates a speedup-based hardware-efficiency factor that captures the effects of tensor parallelism and fully sharded data parallelism. We derive a scaling law model that accurately predicts training energy across heterogeneous configurations.

Scaling reinforcement learning for visual mathematical reasoning requires more than generating harder questions: as data volume grows, the reward labels themselves must remain reliable. Yet existing data pipelines scale supervision while trusting the labeller, and policy-side methods assume the underlying answers are already correct. We instead treat scaling as a verifiable data-construction problem and decouple two axes before any policy update: prompt difficulty, expanded by route-specific evolution operators, and answer reliability, enforced by offline hypothesis-test falsification. We instantiate this as VeriEvol, an iterative framework with two extensible components: a type-aware evolution module that rewrites low-difficulty image-question seeds into harder, image-grounded prompts; and HTV-Agent, a verifier that accepts an answer only after multi-source counter-evidence has failed to refute it. The resulting verified data scales in volume, extends by adding evolution routes or verifier channels, and plugs directly into existing GRPO-style RL recipes. On a five-benchmark visual-math suite, scaling evolved SFT data from 10K to 250K samples raises the mean accuracy from 35.42 to 54.73; then, with backbone, SFT initialization, and GRPO recipe held fixed, VeriEvol adds a cumulative +3.88 over an un-evolved RL baseline, of which +1.82 comes from evolved prompts and +2.06 from the HTV-Agent verifier. We release the prompts, data, models, code, and the full verifier trace of every sample, so that downstream work can scale and audit the pipeline rather than only inspect its outputs.

Text-to-SQL enables users to access relational databases via natural language, but real-world settings remain challenging due to coordinated reasoning over complex database environments. Existing systems often use multi-stage pipelines or reasoning models specialized for individual stages. However, fixed pipelines rely on predefined stage orders, limiting their adaptivity to query demands and intermediate evidence. Recent orchestration-based methods provide flexibility by composing specialized modules for each query, but typical plan-then-execute approaches still commit to a complete workflow before execution and cannot adapt to intermediate artifacts and feedback. In this paper, we propose SQLConductor, a step-wise orchestration learning framework for Text-to-SQL. SQLConductor formulates Text-to-SQL subtasks as specialized actions for workflow composition and trains a policy model to select the next action based on intermediate artifacts and feedback. To learn this policy, SQLConductor introduces Search-to-Policy Learning, which uses Monte Carlo Tree Search to explore candidate workflows and stability estimation to identify robust supervision. The policy model is trained with Stability-weighted Supervised Fine-tuning to prioritize high-quality orchestration patterns and further enhanced through Curriculum Reinforcement Learning. This transforms offline workflow search into a deployable policy for step-wise orchestration at inference time. Experiments on BIRD-Dev and out-of-distribution datasets show that SQLConductor achieves superior execution accuracy and strong generalization, reaching 73.2% EX on BIRD-Dev with a compact orchestration policy coordinating frozen larger action models, outperforming prior methods that directly train comparable or larger Text-to-SQL backbones. Further analyses show that the learned policy adapts orchestration to diverse query demands.

Recent large language models (LLMs) are good at general text generation, but it is still hard to use them for domain-specific data generation because the output must follow strict formatting and structural rules. Unlike open-ended tasks such as question answering or translation, domain-specific generation must be both semantically correct and compliant with existing guidelines and standards. In this work, we study the nationwide interoperability problem of utility power outage reports in the United States. In practice, outage reports need to be machine-readable (e.g., JSON or XML) and must strictly follow requirements from energy-sector regulatory bodies. To address this problem, we propose POTracker, an optimized LLM for power outage report generation. We fine-tune Qwen2.5-7B-Instruct using our proposed objective. The key contribution is a new loss function, POTrackerLoss, that considers both textual similarity and structural (tag) similarity between the generated report and the ground-truth report. We evaluate POTracker on a dataset of 1,000 power outage reports and compare it with five well-known fine-tuning methods and one rule-based XML conversion method. Results show that POTracker outperforms other fine-tuning approaches, improving overall accuracy by up to 51% and reaching 86.47% structural accuracy for generated power outage reports. In addition, we conduct a human study to assess the quality of the ground-truth standard reports, where domain experts assign the generated labels an average score of 4.03 on a 0–5 scale.

Autonomous Underwater Vehicles (AUVs) rely heavily on the fusion of inertial sensors and Doppler velocity logs (DVLs) for navigation. In standard autonomous navigation systems, the DVL measures four beam velocities, thereby enabling the estimation of the AUV velocity vector. However, during real-world missions, the DVL may receive noisy or incomplete beam measurements due to marine obstacles, seabed reflections, or environmental disturbances. Furthermore, some low-cost underwater platforms operate without inertial sensors to reduce system complexity and cost. In such cases, reliable estimation of the AUV velocity vector in real-world missing beam scenarios becomes challenging, leading to degraded navigation solutions. To circumvent these challenges and enable resilient underwater navigation, we propose DVL-DeepONet, a physics-guided deep neural operator framework along with three variants. The proposed models are designed to estimate DVL-based velocity information under multiple operational scenarios, including (i) noise-resilient estimation in coupled inertial/DVL measurements, (ii) DVL-only learning, and (iii) beam measurement recovery. By learning a nonlinear operator that maps temporal inertial/DVL observations directly to vehicle velocity while enforcing DVL measurement physics through a consistency constraint, the proposed approach enables robust velocity estimation even under degraded sensing conditions. The proposed framework is validated using real-world AUV experiments, comprising a cumulative path length of approximately 10,000 m. Experimental results demonstrate that the proposed DVL-DeepONet architectures outperform baseline model-based approaches and learning-based algorithms by 40%.

Medical vision-language models (VLMs) such as BiomedCLIP generalize broadly, but adapting them to a clinical service is as much a safety problem as an accuracy one. Updating a deployed model for a new imaging modality can fail silently in two ways that harm patients: it can forget modalities it already handled (catastrophic forgetting), and it can drift from its trustworthy pretrained prior toward modality-specific shortcuts. We study parameter-efficient continual adaptation through these two properties rather than leaderboard accuracy, presenting CADRE: a frozen-backbone framework combining low-rank adaptation (LoRA) with an online, self-scaling, similarity-aware elastic weight consolidation term that bounds retained-competence loss, and an anchor-to-prior penalty bounding embedding drift from the frozen prior. Two short guarantees, a bound on total consolidation mass and a scale-invariance property, remove the scale-related sources of vanilla EWC's order fragility. Using breast cancer across three maximally dissimilar modalities (histopathology, ultrasound, chest radiography) as a controlled cross-modality stress test, under a multi-seed, multi-order protocol with paired significance testing and training approximately 0.23% of parameters, CADRE attains the highest accuracy, SPQ, and backward transfer and the lowest forgetting among adapting methods, reducing forgetting roughly sevenfold versus the strongest regularized baseline (0.075 to 0.011; paired p=0.023) and achieving positive backward transfer where every baseline is negative. We frame these as stability properties aligned with clinical-safety desiderata, not a deployment guarantee; robustness to distribution shift and adversarial inputs is out of scope.

Scientists do not, by profession, wage war. Yet warfare's vocabulary consistently appears in their abstracts. To quantify the extent to which warfare's vocabulary pervades scientific abstracts, we analyze 21.4 million papers (2010-2025; OpenAlex, PubMed). We additionally run a within-subject war-framing experiment ($N = 801$; 32{,}040 trials) designed to provide causal insight into the effects of militaristic language on persuasion. Between 2010 and 2025, the presence of militaristic terms in scientific abstracts rose 48% in OpenAlex and 32% in PubMed, with the rise accelerating sharply after 2019 (cross-database $r = 0.96$, $p < 10^{-8}$). The prevalence of militaristic language is conflict-aligned at both country and annual scales (Uppsala Conflict Data Program; $r = 0.77$-$0.84$), with the abstracts from the Global South displaying the fastest rise in militaristic language. Among disciplines, social sciences leads in level of such language while engineering and computer science lead in growth. The COVID and post-2022 large-language-model eras also accelerated the rise and narrowed the language gap between native-English-speaking and non-English-speaking countries. In our follow-up experiment, we found that war framing reduced credibility (mean shift $-0.176$ Likert units, 95% CI $[-0.21, -0.14]$; $d_z = -0.28$, $p < 10^{-20}$), funding willingness ($d_z = -0.12$) and policy support ($d_z = -0.08$), with a trend-level increase in sense of urgency ($d_z = +0.07$). Collectively, findings reveal that while scientific abstracts drift toward warfare, the use of militaristic language may erode credibility, funding willingness, and policy support.

AI agents are driving a new software paradigm, with the ability to autonomously call tools, extract information, manage memory, and complete tasks that span applications and data sources. Most existing end-user operating systems, however, are designed for application-centric workflows and offer little native support for AI agents. This mismatch limits the wider adoption of agents and leads to execution overhead and safety risks when running agents on conventional systems. While the concept of agent-native operating systems is emerging, the research community lacks an open testbed to explore the architectural primitives desired for agent-mediated interaction. We present AOHP (Android Open Harness Project), an OS-level agent harness built on the Android Open Source Project (AOSP). The core design principle of AOHP is to treat agents as first-class OS actors, enabling adaptive user interfaces and agent-friendly runtime environments. AOHP preserves the mature Android software and hardware ecosystem while introducing three agent-oriented system mechanisms: personalized service composition, efficient agent interfaces, and secure information flow. Based on preliminary experiments on challenging tasks covering key capabilities of OS agents, AOHP shows clear advantages in task completion (+21.12% completion rate), execution cost (-51.55% token cost), and security-policy compliance.

Chemical language models (cLMs) are widely assumed to learn surface-level syntactic patterns rather than learning meaningful molecular semantics. Here, we apply sparse autoencoders (SAEs) to MolFormer, an encoder-only cLM, to mechanistically examine how molecular representations are built across layers. We discover that early layers rely on position-tracking latents to parse molecular grammar, while later layers encode atom-in-substructure and pharmacologically relevant features. Additionally, we show that non-canonical SMILES produce more disruptive representation shifts than invalid SMILES, driven by position-latent disruption propagating across layers. To support further exploration, we develop InterMol, an interactive visualizer for SAE activations on molecular strings and structures.

Plant leaf disease classification is crucial for crop protection and precision agriculture but remains challenging under complex backgrounds, illumination variations, and severe class imbalance. Moreover, single-architecture models often fail to effectively capture both local and global representations. To address these challenges, this study proposes an adaptive soft Mixture-of-Experts (MoE) framework with cross-architectural routing that integrates EfficientNet-B0, DenseNet-121, and Swin-Tiny to exploit complementary multi-scale, local, and global features. A soft gating mechanism dynamically assigns input-dependent expert weights, while a two-stage refinement training strategy improves optimization stability and generalization. Experiments on a highly imbalanced potato leaf disease dataset achieve 91.68% recall and 92.62% F1-score, surpassing the strongest individual expert by 5.91% and 5.03%, respectively. Additional evaluations on durian and sesame leaf disease datasets yield F1-scores of 94.03% and 97.04%, demonstrating robust cross-dataset generalization and the potential of the proposed framework for reliable real-world crop health monitoring

Learning to decompose videos into persistent objects is a fundamental challenge in unsupervised object-centric representation learning. Despite recent progress, existing methods struggle to simultaneously achieve accurate object segmentation, consistent identities over time, and reliable foreground-background separation. To address these challenges, we introduce UniSlot (Unified Slots), an unsupervised framework for learning robust and disentangled object-centric representations from videos. UniSlot explicitly separates object appearance from its 3D-aware geometric pose in the scene, linking object identity to appearance while leveraging depth to better distinguish objects from their surroundings. UniSlot achieves state-of-the-art performance in unsupervised object-centric video decomposition and tracking across synthetic and real-world benchmarks, yielding substantially tighter object masks and reducing background leakage while preserving object identities. Beyond decomposition and tracking, these improved representations translate directly to downstream tasks such as unsupervised object dynamics prediction, enabling more accurate forecasting of future object trajectories.

Autoregressive decoding with LLMs is primarily bottlenecked by GPU memory bandwidth, especially in edge-computing settings. While quantization is essential for mitigating this bottleneck, most existing methods treat inference as a uniform process and fail to account for the asymmetry between the compute-bound prefill stage and the memory-bound decoding stage. We propose GRINQH (GRaded INput-based Quantization Hierarchy), a weight-only post-training quantization framework that accelerates decoding by unifying quantization and sparsification. GRINQH leverages activation magnitudes as a proxy for computational importance to dynamically assign weight channels to different precision levels, enabling flexible average bit widths during decoding. Evaluated on Llama3 and Qwen3 models, GRINQH outperforms state-of-the-art fixed- and mixed-precision baselines at comparable 3- and 4-bit settings, even enabling effective 2-bit generation. We experimentally verify theoretical speedups by leveraging a hierarchical nested memory layout for multi-precision storage in a custom GPU kernel. Ultimately, GRINQH establishes a new state-of-the-art Pareto frontier for LLM generation, enabling a dynamic trade-off between generation quality and inference speed.

This paper examines the two concepts of digital humanism and evolutionary design. The aim is to identify and highlight potential common structures, synergies, and challenges. How should and can technical systems be designed, and what implications does this have for the design of our environment? In light of the current debate surrounding artificial intelligence, this paper aims to serve as a preliminary study to help better understand the two concepts of digital humanism and evolutionary design within the context of human-centered technological development. Following a brief introduction, the two concepts of Digital Humanism and Evolutionary Design are presented and graphically visualized. The terms of freedom and responsibility in human decision-making, conviviality, and subjectivity are discussed, along with examples illustrating the distinction between human and artificial intelligence (Turing Test and Chinese Room). The various concepts of evolutionary design (e.g., co-evolutionary or sustainable software development, clean code, or green IT) and Gilbert Simondon's concept of the "open machine" are introduced. The interdependencies between functional specialization and open technology development are highlighted. Both concepts share similar structures. In joint cooperation, they can lead to positive effects and mutual synergies. Significant differences lie in the areas of autonomy and determination in decision-making, as well as in genuine and simulated subjectivity. Open technology development is also currently suffering from the functional specialization of software and AI applications due to a purely market- and consumer-oriented approach. Even optimizations for energy efficiency in sustainable software development lead to greater specialization and thus also have a detrimental effect on open and quality-oriented technology development.

LLM agents increasingly load skills, file-based packages of natural-language instructions written by third parties and distributed through marketplaces, that execute with the user's privileges. A single malicious skill can exfiltrate data, hijack the agent, or persist as a supply-chain foothold, which turns the skill marketplace into a new attack surface for agentic systems. Prompt-injection defenses do not carry over to this setting. They rely on a boundary between trusted instructions and untrusted data, whereas a skill is itself a body of instructions, so an injected command sits among many legitimate ones and inherits their authority. We present Locate-and-Judge, a two-stage detector designed for this regime. A lightweight locator scores the structural spans of a skill by the instruction-following attention each span draws and retains only the top-K. A judge then examines the retained spans in detail. Concentrating the costly judgment on a few high-attention spans lets the detector audit an entire marketplace instead of a sample. Compared to direct LLM-based scanning, this approach offers an order-of-magnitude cost reduction, dramatically increasing its scalability at a small cost to recall, and it dominates keyword and regex baselines at comparable expense. Deployed at marketplace scale and at negligible cost, Locate-and-Judge flags skills with high precision, the majority of which we manually confirmed as malicious, surfacing dozens of live malicious skills, including several disguised as benign functionality and many that SkillSpector and Cisco Skill Scanner fail to detect. We release the resulting labeled dataset.

Bias in natural language remains a persistent challenge in both human-written and AI-generated content, affecting domains such as journalism, education, and AI research. Most existing detection methods identify only the presence of bias, with limited support for granular detection, interpretable explanations, neutral rewriting, and openly available trained models. We present UnBias-Plus, an open-source toolkit unifying (1) segment-level multi-class bias classification, (2) biased span localization, (3) neutral text rewriting, and (4) reasoning for each decision. Available via Python, CLI, REST API, and web interfaces, UnBias-Plus supports accessible bias analysis. The toolkit, source code, models, datasets, and documentation are publicly available.

We present HyperQuant (Hadamard, optimallY Packing, Entropy Rice-coding), a unified post-training quantization pipeline for the weights and the KV cache of large language and diffusion transformers. Across a suite of self-contained experiments (Table 1), HyperQuant outperforms the recent HIGGS scheme at every operating point from 3 to 5 bits per scalar (bps) on weights, and beats both TurboQuant and OCTOPUS on KV quantization down to 1.7 bps. Beyond the LLM setting, HyperQuant quantizes the 19B-parameter LTX-2 DiT video model with no observable per-frame artifacts. End-to-end on an H100 at 4 bps, HyperQuant compresses the linear weights ~3.9x and the KV cache ~3.79x at near-lossless quality. HyperQuant combines four known ideas into a single construction: (i) a per-tile Randomized Hadamard Transform that makes the per-coordinate distribution of weights and activations approximately Gaussian; (ii) quantization to a low-dimensional optimal lattice (E8, D4, A2, or Z); (iii) lossless bit-stripping and near-entropy-optimal variable-length Rice coding of the lattice indices; and (iv) bias-correction methods for the KV cache that keep the reconstruction unbiased under inner products, preserving attention semantics. We further integrate the pipeline with 8-bit and 4-bit Tensor-Core MMA paths (fp8-e4m3, int8, nvfp4, mxfp4), and find that int8 beats fp8 on the post-RHT lattice output. Project page: https://moonmath.ai/hyperquant/

The emergence of Large Reasoning Models has introduced exceptionally long Chain-of-Thought traces, creating a transparency burden where critical logic is often buried under massive procedural text. To address this, we present ReasoningLens, an open-source framework designed for the hierarchical visualization and diagnostic auditing of complex reasoning chains. ReasoningLens addresses information necropsy by: (1) structuring traces into interactive hierarchies that separate high-level strategy from low-level execution; (2) leveraging an agentic auditor for automated error detection and tool-augmented verification; and (3) synthesizing systemic reasoning profiles to reveal model-specific blind spots. By transforming unstructured walls of text into actionable insights, ReasoningLens provides a modular foundation for interpreting, debugging, and optimizing the next generation of reasoning-centric AI.

As agentic LLM systems move from prototypes to deployment, evaluating them has become more important and more difficult. Evaluation goals are often left implicit. Without a clear account of what a system is expected to do, how it can fail, and which failures matter, metric choices become difficult to justify, interpret, or validate. We present Litmus, a zero-label system that designs evaluation and monitoring metrics for AI pipelines by eliciting evaluation intent from source code and targeted interrogation. Litmus first identifies what must be measured and why, then converts those answers into constraints for a justified, per-stage metric portfolio. We evaluate Litmus on three real, code-defined pipelines against AutoMetrics and three DynamicRubric baselines. Litmus achieves broader concern coverage on account grouping at matched size, spans more stages, produces a near-zero-redundancy portfolio, and, under a zero-label design-time constraint, ranks first in validity against per-row quality labels on all three pipelines, decisively on scientific QA ($\rho{=}0.72$ vs. $0.47$ for the strongest baseline on the stitched single-vector score), with overlapping intervals on the other two. Results support a shift from automatic metric implementation to automatic metric specification: before asking which metric to compute, evaluation systems should ask what must be measured and why.

SysML v2's textual syntax enables compiler-based validation of model structure and language conformance. However, semantic mistakes that preserve syntactic validity but violate domain rules cannot be detected through compilers. These errors can propagate through the design process and surface late as costly integration failures. This paper presents a human-in-the-loop framework for identifying and repairing such errors automatically. It combines a fine-tuned Small Language Model (SLM) with a domain knowledge graph encoding physical compatibility rules between system elements. The knowledge graph also guides the generation of synthetic training data by systematically introducing plausible domain violations, and augments the model at inference time to ground repair suggestions in valid engineering constraints. We demonstrate the framework using the vehicle systems domain, where the knowledge graph captures the relationships between the mechanical, electrical, fluid, and signal interfaces. Two SLMs, Qwen2.5-Coder-1.5B and DeepSeek-Coder-6.7B, are fine-tuned to output unified diff patches that localize faults and present candidate repairs for engineer review, preserving human judgment in the design process. Evaluation of 1,184 test samples shows that fine-tuning improves semantic fault repair from less than 3% to more than 91%, with patch-based output reducing token length by over 60%. The framework offers a practical path toward AI-assisted model verification that complements existing MBSE tools.

Time series forecasting is a fundamental machine learning task. Recent work has explored Large Language Models (LLMs) for this purpose due to their strong generalization, pattern recognition, and zero-shot or few-shot capabilities. Despite their suitability for long-context learning, LLMs face challenges in multimodal settings: they lack calibrated probabilistic modeling for non-text data and struggle to align heterogeneous representations. To address these issues, we propose a new framework Diffusion-LLM that integrates a conditional diffusion model into an LLM-based forecasting pipeline. This joint design enables learning the conditional distribution of future data while improving semantic alignment in a shared latent space. We evaluate Diffusion-LLM on six long-term forecasting benchmarks, including ETT, Weather, and ECL. Our method consistently outperforms existing LLM-based baseline, achieving notable gains in ultra-long-term and few-shot forecasting and demonstrating the value of distribution-aware regularization for enhancing robustness and generalization in time series LLMs.

Transformer language models have become established tools for modeling human sentence processing, with measures such as surprisal and attention entropy serving as effective predictors of reading difficulty that together capture complementary aspects of processing load. Here, we explore a related class of transformer models: energy-based transformers, which provide a principled formal link to associative memory models, bringing processing research into direct contact with the broader literature on Hopfield networks and dense associative memory. To our knowledge, this is the first exploration of an energy-based transformer measure in computational psycholinguistics. Across reading-time corpora (Natural Stories, UCL eye-tracking, UCL self-paced reading), the energy measure is a robust predictor of reading times, providing significant fit beyond surprisal and entropy in all three. In a controlled experiment on relative clause processing, energy at a single layer captures the well-known object/subject asymmetry. We find evidence that it subsumes effects attributable to both attention entropy and surprisal, suggesting that energy may serve as a single unified predictor where multiple complementary measures have previously been required.

While the wider applicability of LLMs in the legal field is currently debated due to their reliability and the gravity of any errors, narrow uses with well-understood and mitigated risks have emerged. Notably the Swiss Federal Supreme Court uses small on-premises models for tentative translations and short-passage summarization across the four official languages. However, such usage is challenging in the context of Criminal Law. Since rulings and cases employees work on routinely can contain detailed descriptions of violent and sexual offenses, their legitimate work is compromised by refusals and disclaimers due to the activation of model guardrails (over-alignment). To measure this phenomenon, we introduce TF-RefusalBench, a multilingual benchmark for criminal-law translation and summarization derived from public Swiss Supreme Court rulings. TF-RefusalBench contains 5,200 total prompts across French, German, Italian, and English, corresponding to common task prompts and passages likely to trigger refusal. We then use TF-RefusalBench to show that over-alignment is a multifaceted phenomenon, influenced by the model and the prompt and text languages being processed, and that its impact cannot be evaluated solely from an over-refusal perspective, given the disclaimer's impact on task faithfulness. Finally, we evaluate approaches to enable on-premises LLMs for Criminal Law Tasks, demonstrating that while prompting can be effective, abliteration (refusal directions ablation) eliminates refusal with minimal impact on task performance.

Backdoor attacks on molecular graph neural networks (GNNs) are typically evaluated as abstract graph edits, but real molecular learning pipelines do not train on arbitrary graphs. Molecular records must first survive parsing, sanitization, canonicalization, and graph-string consistency checks. We formalize this overlooked admission stage as ChemGuard, an operational protocol for testing whether a submitted molecular record can enter a realistic learning pipeline, while complementing existing defenses. ChemGuard admits a record only when its molecular string is sanitizable and the graph reconstructed from that string matches the submitted molecular graph. Under this operational view, many existing graph-based backdoors lose much of their apparent efficacy because their poisons are chemically invalid or representation-inconsistent. We then show that admission checks alone are insufficient to rule out molecular backdoors. We propose ChemBack, an admission-aware molecular backdoor attack that constructs chemically feasible motif-anchor attachments and ranks admitted candidates by fingerprint-based Tanimoto similarity to clean target-class molecules. ChemBack is model-free during trigger selection, using molecular structures, target labels, fingerprints, and public validity checks, but no victim model, surrogate GNN, learned embedding, gradient, logit, or training-code access. Across molecular benchmarks, validators, architectures, and defenses, ChemBack achieves high attack success with fully admitted poisons while preserving clean accuracy. Our results reveal a two-sided lesson, chemistry-aware admission suppresses many graph-only backdoors, yet chemically valid and target-aligned molecular backdoors remain a practical threat.

Physics-informed neural networks (PINNs) provide an effective way to solve partial differential equations (PDEs) by embedding physical principles into the learning process. However, the conventional PINN formulation, in which all constraints are imposed as soft penalty terms within a composite loss, often exhibits slow convergence, sensitivity to loss weight scaling, and inaccurate boundary enforcement due to poor conditioning of the optimization landscape. To address these limitations, this study proposes a unified hard–soft physics–informed neural network (HSPINN) with adaptive loss weighting. In this framework, Dirichlet and periodic boundary conditions are enforced exactly by construction through analytical or polynomial lifting, masking functions, and periodic feature mappings, while the governing PDE residuals, Neumann fluxes, and initial conditions are treated as soft constraints. An inverse-share softmax strategy dynamically balances the relative importance of individual loss components during training, eliminating manual penalty tuning and improving gradient stability. This formulation ensures boundary admissibility throughout optimization and enhances convergence efficiency and numerical robustness. Applications to representative elliptic (Poisson), parabolic (Burgers), and hyperbolic (convection with periodic boundaries) problems demonstrate that HSPINN consistently achieves faster convergence, higher accuracy, and greater stability than conventional PINNs, establishing a general and scalable foundation for physics-constrained deep learning across science and technology.