Papers

Large language models (LLMs) generate fluent outputs that can be wrong. Unlike humans, who often exhibit cues when providing false information, LLMs produce errors that are difficult to detect because autoregressive decoding provides no mechanism for verifying intermediate reasoning before state progression. We introduce Heaviside Continuity of Rolling Coefficients (HCRC), a verification-first execution framework that reformulates inference as predicate-gated state transitions governed by a Heaviside Gate. HCRC combines model confidence with independent verification signals from a parallel worker architecture, allowing execution to advance only when predefined correctness predicates are satisfied. This prevents invalid intermediate states from propagating, reducing epistemic entropy without modifying the underlying model. We evaluate HCRC on software-engineering and reasoning tasks across thirteen proposers from four providers. On capable proposers, the gate reduces the false-completion rate (FCR) from 4–7% to 0% while remaining latency-competitive and, in some settings, faster than the unwrapped model. On weaker proposers, it converts false completions into honest halts instead of corrupting downstream state. Beyond benchmarking, HCRC has operated for months as the production control plane of an agentic coding environment, authorizing file mutations, verification-driven progress reporting, and memory compaction. These results establish HCRC as a general framework for verification-driven LLM execution, showing that reliable reasoning can be achieved through principled execution control rather than model scale alone.

Automated detection of interictal epileptiform discharges in scalp electroencephalography (EEG) is clinically important, but recent high-performing deep-learning models often trade interpretability for accuracy. We introduce EEG-SpikeAgent, a closed-loop program-synthesis framework that uses a large language model (LLM) agentic system to generate signal-processing features for spike detection in scalp EEG. The system iteratively proposes one deterministic EEG feature module at a time, executes the resulting code on EEG to generate tabular features, evaluates performance via a tabular classifier, summarizes run-level metrics, and feeds structured diagnostics back to the model for refinement. Across iterations, EEG-SpikeAgent proposes and refines candidate signal features and decision rules informed by model performance. We evaluated EEG-SpikeAgent on VEPISET, a public 29-channel dataset of 4-second epochs containing 2,516 discharge-containing and 22,933 non-discharge epochs. Across five-fold cross-validation with a gradient-boosted tree classifier, agent-generated features achieved an area under the receiver operating characteristic curve of 0.935, balanced accuracy of 0.699, F1 score of 0.557, sensitivity of 0.401, and specificity of 0.996 at the default operating point. At an operating point with sensitivity 0.80, mean precision was 0.470 and mean specificity was 0.900. Artifact-aware feature generation improved balanced accuracy and F1 score over spike-only feature search. These results indicate that LLM-based program synthesis can automate EEG feature engineering in auditable and inspectable code-driven manner for clinical and methodological review.

Accurate prediction of patient-specific therapeutic response from pre-treatment transcriptomes is hindered by the scarcity of matched clinical response labels and post-treatment molecular profiles. Preclinical transfer-learning models can simulate drug-induced expression changes but are often hard to interpret and unstable, whereas knowledge-graph methods provide mechanistic context yet remain static and fail to capture drug-induced transcriptomic perturbation dynamics. We propose PREDIKTOR, a patient-centered multi-view framework that aligns a personalized network view with a transferable transcriptomic perturbation view to predict clinical drug response. For each patient, we construct an individualized gene regulatory network from tumor expression using DysRegNet and augment it with drug-target links from DrugBank; a graph neural encoder yields a drug-centric, mechanistically grounded embedding. In parallel, a frozen condition-specific gene-gene attention model pretrained on LINCS L1000 generates a simulated post-perturbation transcriptomic profile for the same patient-drug pair. We align the two views in a shared latent space via a CLIP-style contrastive objective with drug-context hard negatives, then concatenate the representations for end-to-end response classification. On TCGA, PREDIKTOR consistently outperforms state-of-the-art baselines under patient-, drug-, and tissue-split evaluations, and transfers zero-shot to the I-SPY2 trial, improving AUROC by 5.6% over competing methods. The aligned embeddings yield stable gene and pathway attributions that recover known mechanisms, supporting actionable and interpretable precision oncology.

This paper studies a finite-horizon multi-item capacitated lot-sizing problem in which demand quantities are deterministic, while demand-arrival periods are stochastic. Each demand occurs once within a known time window and must be satisfied no later than its deadline. The proposed model makes production and allocation decisions at the demand level, allowing it to represent capacity competition, demand-specific backlog, and allocation-dependent inventory dynamics. The stochastic problem is formulated as a discrete-time Markov decision process (DTMDP), including the state space, feasible actions, transition kernel, and one-period cost function. To isolate the computational effect of stochastic timing, each stochastic instance is first compared with a deterministic counterpart in which each arrival distribution is replaced by its most likely arrival period. This comparison shows that stochastic timing substantially increases the number of states, the number of transitions, solution time, and memory pressure. A genetic algorithm (GA) is then proposed for the stochastic-timing problem. The GA searches over feasible state-feedback policies and evaluates each policy exactly under the DTMDP transition model. Computational experiments on 330 benchmark instances show that the GA remains close to the exact stochastic solution whenever the latter is available, with an average optimality gap of about $3.44\%$. On the difficult benchmark instances, comprising 90 test cases, the GA remains below the $5\%$ optimality-gap threshold and achieves an average optimization speedup of $6.89 \pm 1.41$ at the $95\%$ confidence level. For instances that cannot be solved exactly on the available hardware, an empirical Bellman-time regression is used to estimate the missing exact resolution time and extrapolate the expected GA speedup.

Agentic memory under a fixed budget involves two stages: retention and retrieval. Existing retrieval-centered paradigms implicitly assume necessary evidence survives eviction, but we challenge this by isolating a pre-retrieval failure mode: structurally indirect prerequisite eviction, in which upstream blocks weakly aligned with the query are discarded under budget pressure. We provide an operational definition of this failure, a reproducible deterministic benchmark, and per-seed trace diagnostics. Finally, we evaluate Dependency-aware Semantic Garbage Collection (DSGC), a one-hop graph-aware rule. In our main suite, DSGC improves full-chain retention from 0.03 to 0.90 under a lexical encoder and from 0.23 to 1.00 under a sentence encoder. Robustness checks then identify the budget and scaling regimes where the one-hop rule holds or degrades. Our released pipeline and failure postmortem support mechanistic analysis of retention before retrieval as a distinct failure boundary.

Recent machine learning (ML) advances have demonstrated that deep learning (DL) achieves impressive results in different application domains, including the classification of computer network traffic to corresponding applications. However, the data frequently contains diverging patterns within a single predicted class. This presents a significant challenge to the ability to provide a clear and comprehensive explanation and emphasizes the necessity for tools capable of detecting and analyzing these patterns. Furthermore, the capacity to extract descriptive rules for classes is a crucial requirement in network traffic analysis and intrusion detection, particularly when leveraging advanced tools like next-generation firewalls. We provide a visual-interactive system that explains predictions of classes for network traffic. Global explanations derived from multiple samples of a given class contribute to understanding model predictions. Visualization of global explanations enables recognition of different patterns that offer experts a more comprehensive overview of its characteristics. We introduce a prototype that facilitates visual exploration and refinement of global explanations, enabling network experts to detect and refine new patterns for specific applications. These explanations support the identification of misleading features and the formulation of new rules for the management of networks. Our approach also aims at enabling ML experts to acquire new insights, including the possibility of separating or merging classes and the development of more accurate and reliable DL models. Our proposed prototype was evaluated by experts in machine learning and network analysis.

Multimodal Continual Instruction Tuning (MCIT) is crucial for adapting Multimodal Large Language Models (MLLMs) to evolving a sequence of downstream tasks. Prior methods mostly utilize Mixture of Experts or expansion merge approach, primarily focusing on catastrophic forgetting, yet they still suffer from negative interference during inference, where newly learned updates overwrite useful prior knowledge and degrade overall performance. To address this, we propose SiGMA (Sign Guided Merging and Adaptation), a simple yet effective framework that mitigates negative interference with two components: sign guided adaptive tuning during training and sign guided merging at inference. Sign guided adaptive tuning reduces collisions with past knowledge and learns the current task with minimal drift, mitigating severe forgetting. Sign guided merging further improves consolidation by selectively scaling salient parameters to preserve and amplify useful task specific knowledge. Experiments on UCIT and DCL benchmarks show that SiGMA significantly reduces negative interference and outperforms state of the art MCIT methods. Our code is available at SiGMA.

A central goal of current Spiking Neural Network (SNN) research is to improve their accuracy toward becoming low-power alternatives to Artificial Neural Networks (ANNs). This work further argues that realizing this ambition requires improving not only accuracy but also robustness, defined as the ability to maintain correct predictions under input perturbations. We identify two key issues in existing SNN methods that undermine robustness. First, binary spiking activations can produce large activation-state changes under small perturbations. Second, the lack of effective weight constraints makes network outputs more sensitive to input variations. To this end, we propose Burst Spiking Neural Networks (BuSNNs), built upon Burst-enhanced Spiking Neurons (BSNs) and a Dynamic Weight Constraint (DWC) mechanism. BSNs incorporate burst firing to provide a graded spiking pattern. This spiking mechanism mitigates perturbation-induced transitions in activation states and thereby enhances robustness. DWC penalizes connection weights based on activation states, effectively reducing weight magnitudes and improving robustness while preserving accuracy. We provide theoretical analyses to support these robustness effects. Experimental results further show that, on smaller-scale benchmarks such as CIFAR-10, BuSNNs outperform both SNN and ANN counterparts in accuracy and robustness. On large-scale ImageNet, BuSNN with the MS ResNet-34 backbone further improves top-1 accuracy and corruption robustness over the corresponding SNN baseline by 3.18% and 2.66%, respectively. Despite using spike-based activations, BuSNNs surpass 4-bit activation-quantized ANN baselines and approach 8-bit ANN baselines on ImageNet. They also preserve SNNs' low-power advantage. This work studies the accuracy-robustness problem in SNNs, advancing their practical viability in robust and energy-efficient applications.

Parameter-efficient fine-tuning enables large language models to adapt to downstream tasks with substantially lower computational and storage cost, and Low-Rank Adaptation (LoRA) is among its most widely used techniques. However, vanilla LoRA assigns a uniform rank to all adapted modules, while existing adaptive methods either incur additional optimization overhead or rely on static weights and local gradients that do not capture task-conditioned representation changes. We propose RSRA, a training-free rank allocator that estimates where adaptation capacity is most needed through forward-only representation sensitivity probing on a small calibration set. Specifically, RSRA uses Spectral Effective Rank to allocate capacity across layers, measures module-wise hidden-state displacement under standardized virtual low-rank updates with the Frechet Distance, and combines both signals through hierarchical normalization to produce a task-aware rank configuration before fine-tuning. Across commonsense reasoning and natural language understanding benchmarks with Qwen3-4B and Mistral-7B, RSRA achieves the highest average performance in all three reported model-benchmark settings and a 1.48x-1.93x speedup in allocation time over the fastest competing pre-allocation method. When integrated with DoRA, LoRA-FA, and PiSSA, RSRA improves 15 of the 18 evaluated combinations and increases the average performance of all three PEFT methods.

Urban rail fare systems may be non-additive: the fare of a single paid journey from an origin to a destination can differ from the sum of fares over multiple legally separated journey legs. This paper presents LegalFarePlan, a fare-transparent route-planning framework that models legal exit-and-reentry operations as explicit, auditable constraints. Given a transit network, fare function, transfer rules, station-level exit/re-entry costs, an extra-time budget, and a split limit, the planner computes explainable route plans over paid journey segments. The artifact implements Dijkstra shortest-time and direct route-planner baselines, a greedy split heuristic, bounded exact label-setting, and Pareto-frontier search. Evaluation uses controlled synthetic data and a 57-station semi-synthetic benchmark with 360 OD pairs. On the semi-synthetic benchmark, bounded exact search identifies positive modeled fare reductions for 71.11% of OD pairs, with mean reduction 3.78 and maximum reduction 9.0 synthetic fare units under a 45-minute extra-time budget. These results demonstrate method behavior and reproducibility; they are not empirical conclusions about MTR or any transit operator.

Large-scale, multi-subject widefield calcium imaging provides unprecedented access to brain-wide cortical dynamics. However, the high dimensionality, complex spatiotemporal structure, and substantial task-irrelevant activity in widefield recordings have largely restricted modeling efforts to single-session analyses, limiting scalability and generalization. While multi-subject pretrained models have been explored for some neural modalities, multi-subject models for widefield calcium imaging have not yet been demonstrated; further, subject-invariant zero-shot behavior decoding remains elusive for multi-subject models across neural modalities more broadly. As a first step toward foundation modeling of widefield data, we introduce WiCAT, a multi-subject model that leverages self-supervised pretraining to both outperform single-session models and enable zero-shot behavior decoding on unseen subjects. WiCAT introduces an atlas-grounded tokenization scheme without session-specific components and learns globally shared spatiotemporal representations. Across multiple widefield datasets, the pretrained model supports lightweight downstream decoding, transfers across subjects, tasks, and datasets, and outperforms baseline models. Notably, the model also achieves robust zero-shot continuous behavior decoding and left-out brain region reconstruction on unseen subjects.

The execution harness is a critical component of LLM agents, shaping how the agent interacts with its environment and organizes multi-step execution. Prior work has shown that improving tools, workflows, and other harness components can substantially improve agent performance. This suggests that the harness can be treated as a learnable control layer that adapts execution decisions from experience, rather than as a predetermined execution configuration. In this work, we introduce HarnessMDP, a learning-based framework that makes harness control itself learnable. HarnessMDP formulates agent execution as a finite-horizon Markov decision process, where the harness learns to select structural execution actions from the evolving agent state. The harness policy is learned offline from collected trajectories using task performance and deployed online to adaptively control new executions. Extensive experiments on coding tasks, $\tau$-bench, AgentBench, and other agent evaluation settings show that HarnessMDP consistently improves execution behavior and final task performance over baseline harnesses. These results demonstrate the effectiveness of HarnessMDP and establish learnable harness control as a practical mechanism for improving LLM agents through better execution. Our code is available at https://github.com/Hik289/Agentic-RL-harness.git.

Deep neural networks often contain substantial hidden-state redundancy, but most compression methods operate directly on weights, neurons, or quantised representations without explicitly characterising the dynamical role of internal states. This paper proposes a controllability-observability framework for empirical state-order reduction of deep neural networks. By viewing a trained network as a depth-indexed nonlinear dynamical system, we construct data-driven reachability, observability, and balanced Gramians from hidden-state snapshots and output Jacobians. The resulting A/B/C tests estimate layer-wise reachable, observable, and jointly reachable–observable ranks. These ranks are then used not only as diagnostic measures of hidden-state redundancy, but also as actual compressed layer widths for realised reduced networks. Experiments on MNIST and CIFAR-10 compare the proposed balanced realisation against projection-based reduction, unstructured pruning, structured pruning, low-rank SVD, dynamic INT8 quantisation, and linear baselines. On MNIST, a four-layer SiLU DNN is reduced from state order 1024 to 277, giving 72.95% state compression and 73.48% parameter compression, while maintaining 95.45% accuracy compared with 96.60% for the full model. On CIFAR-10, a larger SiLU DNN is reduced from state order 4608 to 1339, giving 70.94% state compression and 83.09% parameter compression, while preserving accuracy from 54.45% to 54.44% and reducing CUDA inference latency by approximately 3X. The results show that balanced reachable-observable ranks provide a principled empirical minimal-realisation criterion for designing compact neural architectures with little or no loss in accuracy.

While recent advances in large language models have enabled end-to-end automated manuscript generation, existing systems suffer from three critical deficiencies: (i) generated claims are not deterministically grounded in verifiable literature, (ii) experimental results are frequently fabricated rather than executed, and (iii) there exists no standardized, multi-dimensional framework to assess whether AI-generated manuscripts meet the quality and rigor required for real-world publication. We present Prompt-to-Paper, a multi-agent framework that directly addresses this evaluation gap through three integrated innovations. First, a deterministic retrieval-augmented generation pipeline with section-aware relevance scoring and snowball citation expansion grounds every claim in a verifiable corpus of 60–100 papers. Second, an autonomous coding agent executes real computational biology experiments replacing synthetic outputs with genuine numerical results. Third, an eight-dimensional automated quality scorer, benchmarked with approximate reference statistics from published papers and augmented with explicit hallucination penalties, provides standardized, reproducible quality assessments. The quality-driven improvement loop uses a context-rich reviser that routes each iteration to one of three researcher actions and fires a deep research cycle every ten iterations to re-run experiments and re-manuscript from stronger outputs. We validate the system on five bioinformatics case studies; all five cases compiled submission-formatted PDFs with zero out-of-range citations. The improvement loop raises manuscript quality by an average of +17.96 points on a 0–100 scale (maximum +26.04. As partial external checks, a human reviewer scored the five manuscripts at an average of 7.0 out of 10. Complete manuscripts are produced at approximately 0.31 USD per paper.

This paper explores the "Granularity Paradox" in time-series forecasting, wherein finer temporal disaggregation (e.g., Monthly to Weekly/Daily) improves in-sample diagnostics and dataset size (N), but degrades out-of-sample accuracy due to recursive error compounding over longer horizons (H). Conversely, coarse aggregation (Annual) eliminates recursive error propagation but reduces data available to estimators. We formalize this trade-off and benchmark 10 models - spanning naïve, statistical, machine learning, and deep learning architectures - across six granularities using a 13-year public procurement dataset. The empirical results reveal a non-monotonic threshold structure: recursive autoregressive and seasonal models degrade substantially under high-frequency forecasting (e.g., Holt-Winters reaches a Test R-squared of -151 and TPFE of 425.85% at the Daily grain), while the LSTM traces a U-shaped error curve, worsening from Monthly (19.66%) through Bi-Weekly (35.94%) before overcoming the error propagation penalty at Daily (TPFE of 4.35%, R-squared of 0.66). Linear Regression remains stable across all granularities (16.3-17.0% TPFE), confirming that the paradox is driven by recursive feedback topology, not model complexity. The results demonstrate that standard pointwise metrics (RMSE, MAE) systematically mask cumulative error propagation, and that evaluating forecasts without goal-dependent cumulative metrics produces misleading assessments of model adequacy. We introduce a consensus-dissensus diagnostic comparing the directional behaviour of pointwise metrics against cumulative TPFE across granularities, enabling the identification of models whose standard diagnostics mask systematic error propagation.

Accurate work-zone geometry perception is critical for intelligent transportation systems, and ultra-wideband sensing offers a low-cost approach for infrastructure-aided reconstruction. However, outdoor UWB ranging is often degraded by non-line-of-sight propagation, burst noise, and long-tail errors, which can distort downstream spatial reconstruction. We present GAIA, a geometry-aware, infrastructure-anchored learning framework that couples temporal range modeling with latent anchor-layout estimation and deterministic distance projection. GAIA preserves range denoising as the supervised task while orienting the learned distances toward boundary-consistent reconstruction. We evaluate GAIA on a real-world outdoor UWB dataset with synchronized UWB, GNSS, and IMU measurements, and further test robustness using a real-data-calibrated stress-test simulator. GAIA achieves the lowest overall range MSE and highest polygon IoU among evaluated filtering-based and learning-based baselines, reducing MSE by 18.4% and improving polygon IoU by 15.5% over PoseMLP. These results show that geometry-aware range denoising provides an effective path toward spatially coherent work-zone reconstruction.

We present a bidirectional framework for estimating the energy consumption of text-to-video (T2V) and text-to-video-audio (T2VA) models from architectural first principles and observable generation parameters such as resolution and duration, requiring no access to weights, model size, or implementation details. Forward, it predicts energy from generation parameters and architectural principles; backward, it recovers architectural scaling behavior from observed inference times, with accuracy serving as a criterion for architectural validity. Building on the established compute-bound nature of video diffusion models, we demonstrate that each model's energy profile obeys theoretically derived scaling laws, decomposing into quadratic and linear terms whose coefficients directly reflect the underlying architectural complexity. Validated across six open-source models spanning 8.3B-27B parameters and three GPU configurations, this decomposition achieves below 3% MAPE across all architectures. This approach offers a standardized, empirically and theoretically grounded framework for sustainability benchmarking across T2V models and architectures.

Novel category discovery aims to identify unseen classes from unlabeled data by transferring knowledge from labeled categories, but most existing methods perform discovery in opaque latent feature spaces. As a result, they may separate novel categories accurately while providing little insight into what semantic evidence defines each discovered group. We propose xNCD, an explainable novel category discovery framework that performs both representation-based discovery and pseudo-label assignment directly in a structured semantic concept space. Instead of clustering arbitrary deep features, xNCD learns a label-free concept representation by aligning visual features with vision-language similarity priors from pretrained multimodal models, and then applies a unified labeled-and-unlabeled self-labeling objective over concept-space logits. This design makes each discovered category explainable by construction through stable concept signatures and instance-level concept evidence. Theoretically, we show that routing discovery through a semantic concept bottleneck induces a strict restriction of the feature-space hypothesis class, excluding a large family of unconstrained decision rules and biasing induced partitions toward semantically interpretable concept coordinates. Experiments on CIFAR-10, CIFAR-100, and CUB-200 demonstrate that xNCD preserves strong discovery performance while providing intrinsic explanations. Under task-agnostic evaluation, xNCD achieves 92.63% overall accuracy on CIFAR-10, close to UNO's 93.4%, and improves CIFAR-100 overall accuracy from 73.2% to 76.45%, while being the only compared method that provides human-readable cluster- and instance-level explanations.

Learning action-conditioned world models for dexterous manipulation that are genuinely controllable requires capturing complex, high-dimensional hand kinematics from limited real-world data. We present Mask2Real-WM, a two-stage world model that improves controllability for dexterous hands under a limited real-data budget by decoupling pixel prediction into a dynamics model and a rendering model. The dynamics model predicts future segmentation masks from past masks and a high-dimensional action sequence. The rendering model converts the predicted masks into photorealistic RGB images. This design allows us to train the dynamics model on a large synthetic dataset spanning the full range of hand motions and interactions. We then fine-tune the dynamics model and train the rendering model on only 2.5 h of real demonstrations to obtain a controllable world model. We compare five models with matched training recipes on a 23-DoF robotic system, measuring per-DoF controllability both with random target commands and with sinusoidal per-DoF actuation, judged blind by human raters. The decoupled model achieves the strongest controllability, with further gains from simulation data. Beyond improved controllability, Mask2Real-WM produces sharper frames while maintaining comparable perceptual quality.

Every LLM agent run re-derives its behavior token by token on a frontier model: brilliant, expensive, slow, and unbounded. We present Auto, a compiler that records live agent behavior, measures which parts are secretly deterministic, extracts them into verified programs or distilled specialists, and emits cognition binaries: WebAssembly artifacts whose manifests carry measured guarantees and whose declared capabilities are physically enforced by the sandbox. A tiered runtime executes compiled behavior behind conformally calibrated guards; guard trips deopt to the reference agent, and the captured trace recompiles back down, so nothing is figured out twice. We use "AGI compiler" in one narrow, testable sense: a system that autonomously converts novel experience into permanent, verified, near-free skill while measuring what it does not know. On AUTO-BENCH, a benchmark we introduce and pre-register, 87.1% of 560 recorded frontier-agent spans are witnessed-deterministic (three of the four censused task families measure 100.0%). On a 300-item stream with three scheduled distribution shifts, the closed loop compiles three artifact generations and drives marginal cost from 59 to 2 micro-dollars per item (6.4x end-to-end) at 96.9% parity on witnessed inputs with zero errors. The same stream also quantifies the failure modes: a loose guard silently mislabels 48.9% of compiled answers, and an unfaithful deopt reference causes the verification gate to refuse recompilation. Calibration and reference fidelity, not model capability, decide whether cheap stays correct. Code: https://github.com/RightNow-AI/auto

Camera-radar (CR) fusion is a practical sensing configuration for autonomous driving, but existing models are typically trained with task-specific supervision, limiting reusable representation learning. We present CRISP, a spatiotemporal CR backbone pretrained through forecasting-based representation learning. Given historical multi-view images and radar sweeps, CRISP learns a unified bird's-eye-view (BEV) representation by predicting future LiDAR point clouds. LiDAR is used only as privileged supervision during pretraining; the deployed model requires only camera and radar. To make forecasting-based pretraining effective for CR fusion, CRISP introduces an enhanced radar encoder, radar-enhanced temporal self-attention, and multimodal feature rendering with modality innovation gating. These components inject radar range and Doppler cues into BEV temporal propagation and allow BEV tokens to selectively incorporate camera and radar evidence. Experiments on nuScenes show that CRISP improves long-horizon point cloud forecasting and transfers effectively to downstream tasks, including 3D detection, tracking, online mapping, motion forecasting, future occupancy prediction, and planning, suggesting that predictive CR pretraining is a promising path toward scalable driving representations under practical sensor configurations. The project website is https://umfieldrobotics.github.io/CRISP.

Code language models are now trusted collaborators in production workflows for debugging, refactoring, and iterative repair, and every benchmark that evaluates them assumes the instructions they act on are correct. We study what happens when that assumption breaks. We evaluate code language models across four experiments designed to assess whether models resist or obey incorrect instructions in single-pass and iterative repair settings, using the RunBugRun dataset of algorithmic Python problems with deterministic test cases. Our findings reveal a striking behavioral pattern: models correctly identify an incorrect instruction as wrong, then follow it anyway. This compliance unknowingly introduces errors beyond the original bug, and the corrupted code state cannot be recovered through subsequent self-guided iterative repair, which fails to converge across passes. We term this Blind Obedience, characterize the Ghost (Unknown) Errors it introduces, quantify the proportion of cases where semantic corruption proves irrecoverable, and show that extended reasoning cannot reverse it. These findings surface behavioral properties invisible to pass-rate evaluation, with direct consequences for code language models deployed in production settings.

Low-precision neural networks are attractive for resource-constrained hardware, but fixed-point arithmetic introduces failure modes that are often hidden by idealised quantisation models. In particular, two's-complement overflow wrapping can corrupt hidden activations by changing both their magnitude and sign, leading to unstable numerical error propagation and severe accuracy degradation. This paper proposes a Lyapunov-stabilised quantisation framework for low-precision neural networks operating under hardware-style wrapping arithmetic. The hidden-state energy is monitored through a layerwise Lyapunov function, and a monotone projection is applied to enforce bounded and non-increasing state evolution across depth. The method is evaluated on MNIST using a compact patch-based transformer under post-training quantisation and quantisation-aware training with fixed-point bit-widths from 4 to 16 bits. Monte Carlo results show that unconstrained wrapped quantisation-aware training collapses to near-chance accuracy across 6-16 bits, with activation overflow rates exceeding 11%. In contrast, the proposed monotone Lyapunov projection suppresses activation overflow to below 0.012% and restores stable low-precision learning, achieving 86.55% accuracy at 12 bits. These results demonstrate that Lyapunov-based state control can act as a hardware-aware stabilisation mechanism for reliable fixed-point neural inference and training.

Software agents are becoming increasingly important in modern coding workflows, where they solve complex tasks through multi-step interaction with code repositories. A central component of these systems is the execution harness, which governs how the underlying LLM observes, acts on, and verifies its environment. We show that the harness can in turn shape the agent's evolving beliefs about the environment, even when the task, environment, and base LLM remain fixed. To reveal this effect, we introduce a belief-rollout diagnostic that traces how an agent's anticipated world state and future decisions evolve under alternative harnesses. We further separate immediate divergence introduced by the interface from divergence that develops over longer horizons. Across controlled coding tasks include HIBench-Code, SWE-bench Verified and Terminal-Bench, we find substantial cross-harness belief divergence that is often invisible from terminal success alone and can propagate into different downstream actions. Building on this observation, we introduce Belief-Invariant World-Modeling (BIWM), which exposes evidence hidden or compressed by harness execution and aligns the resulting belief trajectories across different harnesses. Our findings reveal that harness design can systematically shape how software agents understand and act on their environments, providing practical guidance for building and evaluating more reliable coding agents. Our code is available at https://github.com/Hik289/Harness-induce-bias.git.

First-shot anomalous sound detection in DCASE Challenge Task 2 must flag anomalies of unseen machine types with a single threshold, without knowing whether a test clip comes from the data-rich source domain (990 normal training clips) or the data-scarce target domain (10). Two organizer-reported problems remain open: source- and target-domain AUC are negatively correlated across systems, and development-set performance does not predict evaluation-set performance. We address both with a training-free post-hoc layer over frozen audio embeddings: (i) per-domain quantile calibration shrunk toward a pooled map by a prior strength m, tracing a source/target balance frontier, and (ii) a label-free cross-validated domain-balance criterion that ranks candidate configurations from training normals only, paired with a coarse development-labeled viability veto. On DCASE 2025, the criterion rank-predicts the official evaluation score across a 45-configuration grid (Spearman rho = +0.91; family-block bootstrap 95% CI [+0.83, +0.95]) while development score is uninformative (+0.06). Criterion-based selection raises the evaluation score from 55.83 to 59.34 (jackknife CI [2.2, 4.8]) and, on an extended grid, to 61.05 – retrospectively fourth of 35 teams. Replicating on DCASE 2023 and 2024 bounds the claim: development score is uninformative in all three years and degenerate configurations recur (vetoed every time), but under family-clustered uncertainty the criterion's predictive evidence survives only in 2025; in both replication years a fixed full-equalization default matches or beats criterion-based selection. A DCASE 2026 forward test is frozen before the 2026 evaluation ground truth is released; all headline numbers are reproduced by the official evaluator.

How concepts are represented in neural networks is a fundamental question in machine learning. The dominant view treats concept representations as stationary geometric objects. Yet concepts appear in context, and context transforms them. Drawing from neural population geometry, we formalize concept representations as point-cloud manifolds and contextual transformations as vector fields, and instantiate this framework in large language models. Across six model families of varying scales, we find that context moves each concept differently. The variance in these displacements is semantically organized, correlating with lexical concreteness and density. Importantly, both the concepts being transformed and this variance structure are shared across models: displacement structure transported from one model predicts held-out displacements in others significantly above chance. Together, these findings show that models share a common geometry not only in how concepts are represented, but more importantly in how context transforms them, a structure with richer organization than prior work has recognized.

Generic statements like "tigers are striped" and "cars have radios" communicate information that is, in general, true. However, while the first statement is true in principle, the second is true only statistically. People are exquisitely sensitive to this principled-vs-statistical distinction. It has been argued that this ability to distinguish between something being true by virtue of it being a category member versus being true because of mere statistical regularity, is a general property of people's conceptual machinery and cannot itself be learned. We investigate whether the distinction between principled and statistical properties can be learned from language itself. If so, it raises the possibility that language experience can bootstrap core conceptual distinctions and that it is possible to learn sophisticated causal models directly from language. We find that language models are all sensitive to statistical prevalence, but struggle with representing the principled-vs-statistical distinction controlling for prevalence. Until GPT-4, which succeeds.

Vision-Language-Action (VLA) models are commonly treated as end-to-end action policies conditioned on natural-language task descriptions. In practice, however, their behavior often depends sharply on how the instruction is phrased, suggesting that language is not merely a task label but an optimizable conditioning input. We study whether frozen VLA policies can be improved by optimizing language space rather than updating action weights. Our method introduces a language-conditioning space policy that translates a human instruction into a short VLA-grounded command using object appearance, spatial relations, and target-grounding cues. The language-conditioning space policy is optimized with reinforcement learning from sparse task-completion rewards, while the downstream VLA remains fully frozen. Experiments on RL4VLA and VL-Think show that language-conditioning space optimization improves success on instruction-sensitive, symbolic, and multi-object manipulation tasks, demonstrating that language can serve as an optimizable variable for robot foundation models. Website: https://cognitiveaisystems.github.io/VLA-Grounder/

Emergent misalignment (EM) — the broad misbehaviour a language model acquires after fine-tuning on narrow harmful data — is mediated in Qwen2.5 models by a latent persona direction, and that direction is causal in open weights. Transplanting it into a model that shares only pretraining with its source induces broad EM (2.83 $\pm$ 0.26% misaligned against a random-direction floor of $\sim$1.1%), and ablating a model's own direction roughly halves an overt inducer's broadcast (21% to 10%). The transplant doubles as a measurement method, causally assaying directions that a source model represents but cannot itself express. Whether a fine-tune recruits this persona depends on method and capacity, and since low-rank PEFT is the cheaper regime at scale, the recruiting method is also the economical one. On Qwen2.5-32B, LoRA at low ranks on insecure code recruits it (3.4% misaligned) while full SFT on identical data does not (0.3%) and moves against the persona axis (drift–persona cosine $+0.17$ at rank 1 to $-0.10$), the far-inducer, high-capacity exception consistent with a representational-distance $\times$ capacity account. The persona's causal role is itself conditional. Steering a bad-medical SFT run away from the direction during training raises the broadcast from ${\sim}24\%$ to ${\sim}50\%$ while matched random controls stay at or below baseline, replicated across three training seeds, so removing the direction is no blanket recipe. Because recruitment is a loss-reducing shortcut that capacity renders redundant, it can be screened for and prevented in the tested instances. Persona loss-relevance at the SFT solution orders four inducers' broadcasts rank-perfectly within Qwen2.5, inoculation removes recruitment selectively (4.75% to 0.0%, code coherence 65% to 87%), and fine-tuning orthogonal to the single behaviour-derived axis reduces it persona-specifically.

Agentic AI-for-Science can automate ideation, planning, and analysis, but final validation still depends on real experiments. A self-driving lab (SDL) can execute those experiments, yet the loop still has bottlenecks: the agent may spend too many rounds on low-value experiments, or each round may require a high-cost experiment. We target these two physical bottlenecks with one agent. First, a prior-aware agentic DOE loop uses domain knowledge and past results to propose feasible and informative next experiments, reducing trials-to-target. Second, a cost-aware surrogate agent predicts high-cost, high-resolution measurements from low-cost, low-resolution measurements. It chooses between a high- and a low-cost measurement based on the predicted uncertainty. We examine these directions in the biology and materials domains, respectively. Together, under a single agent, these components aim to accelerate the SDL loop by reducing both the number of loops and the cost per experiment.