Code-graph and embedding tools are routinely added to LLM coding agents on the premise that they supply repository structure the model lacks. We report a pre-registered study of how much such tools help off-the-shelf agents localize the code an issue requires changing and how the benefit varies with model size. Nine open-weight models (Qwen3 0.6B to 32B, Gemma 4 E2B to 12B) run one fixed localization harness on 500 held-out SWE-bench tasks under four tool sets (grep only; plus embedding search; plus code-graph tools; both) and a placebo arm whose structural tools return random entities, with five sampled seeds per condition: 170,500 episodes under hypotheses, tests and claim criteria frozen before the runs. The tools raise canonical function Acc@10 by +0.06 to +0.10 for the Qwen3 models up to 14B and the two small Gemma 4 models, by +0.03 at 32B (a near miss after correction), and not at all for Gemma 4 12B, which rarely submits; the gain is larger for small models than for the largest model in both families. The placebo shows that the gain is caused by what the tools return rather than by having them, and that random tool output harms two small models. The pre-registered claim that a 4B model with tools matches a 32B model without them fails. Secondary analyses show that the embedding index, not the code graph, carries the gain; that the gain replicates on tasks created after every model's release and survives a shell harness, thinking mode, FP8 serving and a file-level metric that output formatting cannot inflate; and that it is mostly exposure: the tools show the gold code more often. Injecting the gold code directly explains the size trend. A 32B model already finds the code without the tools, so they add little exposure, while small models act on a shown answer less often (conversion 0.09, 0.24 and 0.40 at 1.7B, 4B and 32B). We release registration, code, episodes and analyses.
Discovering equations from data provides governing laws, constitutive relations and closure models for scientific understanding and prediction. Yet a method that succeeds on one system may fail on another or on noisy data, so discovery depends on expert trial and error. Unsuccessful searches can still reveal useful parts of an equation, but these clues must be recognized and combined. We present LegoSR, a unified framework for agent-guided equation discovery built on composable discovery and cumulative experience. Seven complementary equation discovery methods and shared computational modules form reusable building blocks, connected through a common equation representation. Researchers can launch discovery with one command and guide it with partially known equations. LegoSR Agent selects methods, revises searches using feedback and combines partial findings into new hypotheses. Its Experience Library consolidates graded outcomes into lessons that guide later tasks across equation families. Evaluation covers 134 algebraic regression problems, 40 ordinary differential equation targets and 17 partial differential equation problems at several noise levels, together with three case studies. On noisy differential equation benchmarks, LegoSR Agent nearly matches the recovery coverage of seven expert-tuned methods combined and exceeds a generic agent with the same model, tools and budget by up to 29.4 percentage points, with 51-61% shorter median wall-clock time.
Deep learning approaches for retinal optical coherence tomography (OCT) predominantly rely on predefined disease categories, limiting their ability to detect rare or previously unseen abnormalities. While visual foundation models provide transferable representations, their effectiveness in identifying pathological deviations from normal retinal anatomy remains insufficiently understood. In this work, we introduce an unsupervised framework for retinal OCT anomaly detection and localization that adapts pretrained visual representations to normal retinal anatomy through anatomy-aware reconstruction and contrastive triplet learning guided by synthetic structural perturbations, without requiring disease-specific annotations. We systematically investigate the influence of pretraining strategies, anatomy-aware representation adaptation, and anomaly scoring mechanisms on the detection and localization of unseen retinal abnormalities. Our findings demonstrate that adapting pretrained representations to normal retinal structure improves anomaly sensitivity, while pixel-space reconstruction errors and feature-space deviations provide complementary signals for identifying pathological changes. These findings underscore the importance of anatomy-aware representation learning and offer insights into developing generalizable retinal anomaly detection frameworks beyond predefined disease categories.
The drift-diffusion equations are the fundamental building block for the modeling, design, and optimization of modern semiconductor devices. Solutions of the drift-diffusion equations rarely give way to analytical forms, and exhibit strong nonlinearity, with the solutions ranging over several orders of magnitude. These attributes have led to research into numerical solvers, and while many commercial Technology Computer Aided Design (TCAD) packages now exist, they may require long simulation times, exhibit sensitivity to initial guesses, and rely on mesh generation. Further, these solvers are not well suited for inverse problems where the design parameters of a device are retrieved from its desired functional characteristics. To address all these challenges, this work demonstrates the usage of Physics Informed Neural Networks for the Drift-diffusion (PINND) equations with errors comparable to traditional simulation software using advanced training techniques to solve problems beyond the well-behaved equilibrium regime. For this study, we specifically choose an abrupt PN-junction in both one and two dimensions that features extreme gradients of the carrier concentrations at the Ohmic contacts and an abrupt discontinuity in the doping profile. Using the proposed PINN architecture, we are able to retrieve accurate results for all observables under all bias conditions. Additionally, we apply the PINN solver towards the inverse problem of retrieving doping profiles directly from desired electric field distributions or electrostatic potentials under non-zero applied biases. Our work establishes a powerful PINNs-based approach for semiconductor device modeling with the ability to solve inverse problems under non-equilibrium conditions for microelectronics research and discovery.
Large Language Model (LLM) agents with tool use can behave inconsistently across independent runs from identical states. They may call different tools, produce contradictory outputs, or follow divergent strategies.This variability compounds over time and undermines deployment reliability. Existing evaluation metrics primarily measure task correctness via completion-based scores and do not capture consistency in agent behavior across executions. Separately, agent behavior is analyzed either at the token level using probability distributions or at the single-response level using semantic clustering, but neither captures how inconsistency propagates across future turns of a trajectory. We introduce Recursive Semantic Divergence (RSD), a trajectory-level consistency metric adapted from bisimulation metrics in reinforcement learning. RSD measures the expected cumulative semantic disagreement between two independent rollouts of the same agent from the same state, using natural language inference to detect logical inconsistency. It is defined recursively over future states and approximated by a neural network trained on independent rollouts. The learned metric estimates future divergence from the current state, providing a consistency score at each turn without requiring the full trajectory to complete. When used as a reward signal for fine-tuning, RSD reduces AUC coefficient of variation (CV) by up to 30.9% on tau2-bench and 50.5% on ToolSandbox. Across Qwen-2.5 and Qwen-3.5 models, RSD reduces AUC CV in seven out of eight model-benchmark settings, while improving the final task progress in all eight settings.
We study online convex optimization (OCO) with time-varying movement costs, where the learner incurs both a loss and a time-dependent penalty for changing its decisions. While previous work has established dynamic regret guarantees for this setting, obtaining strongly adaptive guarantees that hold simultaneously over arbitrary intervals remains an open challenge. We address this by developing a new algorithm which, for every interval $[a,b]\subseteq[T]$ and arbitrary comparator sequence, achieves a data-dependent regret bound of $ \widetilde{\mathcal{O}}\left( \sqrt{(D^2+DP_{[a,b]})\sum_{t \in [a,b]}(\|g_t\|^2+\lambda_t\|g_t\|)} \right) $, where $D$ is the diameter of the decision domain, $P_{[a,b]}$ is the path length of the comparator sequence over $[a,b]$, $g_t$ is the gradient of the loss, and $\lambda_t$ is the movement-cost coefficient at round~$t$. Key to our approach is a novel reduction to sleeping experts with movement costs, a previously unexplored variant that is of independent interest. As a result of our development, we further obtain strongly-adaptive dynamic regret guarantees under arbitrary delayed feedback with a refined dependence on interval-restricted cumulative delays. Our approach also naturally extends to OCO with time-varying memory, where we obtain strongly-adaptive guarantees which improve the best-known bounds.
Recent multimodal large language models (MLLMs) can perceive a traffic scene, reason about it, and propose a driving action within a single model. However, existing MLLM-based drivers reason from scratch at each frame and commit to a single explanation, discarding earlier hypotheses about the scene. When the scene is ambiguous, the agent locks in on the wrong interpretation and acts on it before evidence accumulates. In this paper, we propose DriveMind, an MLLM-based driving framework that keeps several hypotheses about the scene alive at the same time and updates their weights as new frames arrive. Inspired by particle filtering, DriveMind propagates each hypothesis through a VLM, scores it against the current observation, and resamples to maintain diversity. On Bench2Drive, adding belief tracking to the same driving pipeline improves the Driving Score (DS) from 68.50 to 81.37 and the Success Rate (SR) from 41.36% to 54.09%. The largest gains occur on ambiguous scenarios such as unsignalized intersections (+20.0 DS) and oncoming-traffic interactions (+20.5 DS). Beyond the overall metrics, DriveMind also improves fine-grained interactive driving abilities such as merging, emergency braking, and give-way.
If agricultural machinery is to support intelligent, offline assistance, then the system must access technical documentation dynamically. In this paper, the Semantic Synchronization of operator manuals over ISO 11783-based CAN buses is analyzed, defined as the process that brings the local knowledge base of a tractor-side assistant into agreement with the operator manual of the connected implement. Data transfer performance was benchmarked using Markdown, JSON, and XML; various open-weight Large Language Models (LLMs) were also evaluated in cross-lingual query tasks. The automatic evaluation of the answers was validated against four human raters. A complete manual was transferred over a physical ISO 11783 network in 31 to 89 s, depending on format and transport settings, which is comparable to the object pool upload that operators already wait for at the first coupling of an implement. Markdown produced the smallest files, while the question-answering performance was comparable across the three formats for most models. With retrieval-augmented generation, locally run models answered most questions on the English manual correctly, with no advantage of the largest model over 7B and 8B models, but performed worse on the five other language editions. Keyword retrieval failed across languages, whereas semantic retrieval provided the cross-lingual capability. The exchange requires only a new parameter group that uses the existing request and transport mechanisms of ISO 11783, which makes it a candidate for standardization. The study is a proof of concept with inference on PC-grade hardware; deployment on embedded agricultural hardware remains future work.
While monolingual language model (LM) surprisal is widely established as a predictor of reading times, it cannot capture the non-native (L2) reading dynamics that are shaped by the reader's first language (L1). In this work, we investigate whether the cross-lingual transfer within bilingual models aligns with human L1-driven biases during L2 reading. We evaluate bilingual models on eye-tracking data of readers from 12 diverse L1 backgrounds reading English texts. To isolate L1-specific transfer, we compare bilingual models trained on an L1 that matches the readers' L1 background to both a monolingual English baseline and bilingual models trained on other L1s. We demonstrate that matching the readers' L1 improves the prediction of L2 reading times. This predictive advantage is concentrated in late-stage eye-tracking measures, modulated by L1-L2 script overlap, and peaks among intermediate-proficiency readers. These findings provide evidence that bilingual LMs capture aspects of cross-linguistic influence, and that such influence is present in L2 reading behavior.
Many questions about a video concern one physical object. Where does the ball end up after the cups are shuffled? Which card is the queen? How often was the mug covered? Open video-language models (Video-LMs) recognise such objects in single frames but do not keep them apart over time. On shuffled shell games, open models from eight families stay near chance, and so do published tool agents. We separate semantic identity binding from temporal identity maintenance. Binding should happen at a moment when the evidence for the object is visible, and maintenance should be left to a tracker. Track, Then Bind implements this principle without training. A promptable video tracker follows every candidate object and draws it with a track number. A frozen vision-language model (VLM) answers one question about the identifying evidence, which binds the target to a track. The answer is then derived from a text description of the tracks. With Qwen3-VL-32B, the agent solves both VET-Bench tasks, where a fine-tuned trajectory model reaches 92% and 90%. On real cup games from the Perception Test, it raises accuracy from 35.5% to 52.3%. Without retuning, the same configuration also solves VET-Bench with Gemma 4 and GPT-5.5. Its track-derived visibility timelines add 8.8-17.9 points on TOC-Bench for open models. GPT-6.1 already answers 92.5% of the real cup games directly but 71.0% through our pipeline, while our timelines still raise its TOC-Bench accuracy from 52.5% to 57.7%. The benefit of external tracking thus depends on the underlying model and the task.
Real-time polyp segmentation in computer-assisted colonoscopy is hampered by the difficulty of deploying high-accuracy models under strict latency budgets. Frame skipping, which performs inference every K frames and reuses or interpolates the rest, improves throughput but introduces motion blur and lesion misalignment that degrade temporal consistency. To close this gap, we propose the Lightweight Frame Compensator (LFC), a 77K-parameter plug-and-play module inserted before a frozen segmentation backbone. Operating on a three-frame window, LFC fuses consecutive frames via spatio-temporal attention into a single compensated frame, and is trained only on pseudo-labels generated by a differently seeded model, without any expert annotation. Experiments on three polyp video benchmarks show that LFC recovers most of the accuracy degraded by direct frame skipping and, in several settings, reaches the accuracy of frame-by-frame inference, while retaining a 2.05X--2.55X throughput gain. We further evaluate the temporal stability of three backbones spanning CNN, pyramid Transformer, and global-attention Transformer architectures, revealing a spectrum of sensitivity to frame skipping that arises from how locally or globally each architecture aggregates misaligned input. The compensation gain of LFC follows this spectrum: it is largest for high-capacity Transformer backbones, which achieve the best static accuracy but suffer the most under skip-frame inference. LFC thus offers a lightweight, pluggable, and deployment-friendly route to real-time polyp segmentation.
Large Language Models (LLMs) can generate UML class diagrams from natural language requirements, but hallucinations may compromise the correctness and completeness of the resulting diagrams. Existing evaluations mostly assess the final diagram, without examining the intermediate traces of reasoning-capable LLMs, which can reveal where these hallucinations arise. We develop an empirical taxonomy of eleven hallucination sub-types organised into three dimensions: Prompt Faithfulness (deviations between requirements and reasoning), Code Faithfulness (discrepancies between reasoning and generated PlantUML code), and Reasoning Quality (problems within the reasoning). The taxonomy was derived from 60 manually annotated DeepSeek-R1 traces, applied to 120 further DeepSeek-R1 traces through a dual-LLM annotation pipeline, and examined on 20 DeepSeek-V4-Pro traces. Code Faithfulness was the most frequent dimension for DeepSeek-R1 in both datasets, whereas Prompt Faithfulness dominated for DeepSeek-V4-Pro, suggesting model-dependent failure patterns. In the manually annotated traces, reasoning-trace length showed no significant monotonic association with hallucination frequency, whereas requirement complexity was negatively associated with it, particularly for Prompt Faithfulness errors. In a mitigation experiment on 13 Golden UML ModelSet cases with three Qwen3-based models, adding Prompt Faithfulness and Reasoning Quality constraints to the generation prompt raised the pooled mean F1 score by 4.0 percentage points, whereas a post-hoc Code Faithfulness review provided no consistent further benefit. Relationship modelling remained the lowest-scoring element throughout. These results suggest that localising hallucinations within the generation process can inform targeted, prompt-based mitigation. Whether these findings generalise beyond the DeepSeek and Qwen3 model families remains to be tested.
Medical knowledge evolves, yet whether large language models (LLMs) faithfully incorporate revisions into clinical recommendations remains unclear. We introduce ChronoEvo, comprising 1,709 recommendation-level revisions across 12 forms of medical knowledge change and 275 guideline lineages, to distinguish recognition of revised guidance from its full incorporation. With updated guideline evidence explicitly provided, revision recognition exceeded full incorporation across 14 models, with a 20.1-percentage-point gap in the strongest model. For Qwen3-4B, 90.8% of responses recognized revisions, 80.2% were clinically correct, and 58.5% fully incorporated them, yielding a 32.3-percentage-point gap. Two clinicians' adjudicated assessment of 300 responses yielded a 40.3-percentage-point gap in this enriched validation sample; 61 of 144 responses with incomplete incorporation were clinically incorrect. Prompting Qwen3-4B to prioritize revised evidence reduced outdated-content retention and improved clinical correctness without significantly improving full incorporation. These findings identify faithful incorporation of revised guidance as a distinct challenge for clinical LLM reliability.
Reasoning LLMs often continue reasoning after reaching a correct answer, wasting computation and sometimes replacing it with an incorrect one. Deciding when to stop is a metareasoning problem: the model must weigh what further reasoning could achieve against its cost, given what it can already answer. Direct utility training learns a stopping representation at a chosen token price; supporting new prices can require additional controller training. We introduce BELT (BELief-based Termination), which separates outcome prediction from stopping control. It learns beliefs about current and future answers from a frozen LLM's hidden states, then fits a linear stopping policy on the frozen features. The prediction targets are independent of token price, so one probe supports different accuracy-cost trade-offs through CPU policy refitting or threshold adjustment. Averaged over five out-of-distribution (OOD) math benchmarks, BELT reduces generated tokens by 41% on Qwen3-1.7B and 51% on OLMo-3-7B-Think without loss of accuracy. The same math-trained BELT transfers to coding and agentic tasks without training on code or agentic data. In all three domains, its accuracy-cost trade-offs are competitive with or better than those of end-to-end controllers. On OLMo3, its probe has 21.2x fewer parameters than one checkpoint of the strongest baseline, OS-Pruner, and needs 55x fewer training GPU-hours than OS-Pruner's five price-specific checkpoints.
We propose FiLM-CAM; a conditional-access image watermarking approach that enforces authorized encoding, decoding, and removal through keyed feature modulation. A secret key conditions a Feature-wise Linear Modulation (FiLM) mechanism that transforms the internal representation of the watermark signal before embedding, thereby coupling the payload to the key at the feature, rather than pixel level. The same key must be presented to condition the decoder to correctly recover the payload embedded in an image. To enforce conditional access, we train the system to maximize the performance gap between authorized and unauthorized use, ensuring reliable extraction under the correct key while inducing failure under incorrect keys. In addition, we design our encoder to estimate the watermark residual invariant to the presence of pre-existing watermarks. This design enables conditional watermark removal via simple re-encoding and subtraction of the residual using iterative refinement, contingent on knowledge of the secret key. We evaluate FiLM-CAM across multiple watermarking encoder-decoder backbones, demonstrating that keyed feature modulation serves as an effective, architecture-agnostic conditional access module (CAM) for image watermarking.
We prove that for every open $U \subseteq \mathbb{R}^n$ the spherical projection of every gradient trajectory $\theta(t)$, $t \geq 0$, of a definable $C^1$-potential $F \colon U \to \mathbb{R}$ has finite length and, in particular, converges (directional convergence). In other words, every gradient trajectory of a definable $C^1$-potential has finite total angular motion. Here, definable means definable in an arbitrary o-minimal structure expanding the real field. Notably, the gradient trajectory in this result is an arbitrary solution of the gradient flow differential equation (whose solutions may not be unique) with no further assumptions: The gradient trajectory may be unbounded and diverge to infinity or it may be convergent. The established theorem therefore provides affirmative answers to both the gradient conjecture at infinity for definable potentials and the definable Thom gradient conjecture. We extend our conclusions from the Euclidean metric to a general class of ambient metrics including Riemannian metrics and we establish uniform local upper bounds for the lengths of the spherical projections of gradient flow trajectories. We finally apply our findings to the training of deep neural networks (DNNs) to conclude that, in the case of definable and continuously differentiable activation and loss functions, the spherical projection of the DNN training flow must always converge (directional convergence).
One-shot landmark localization in volume scans with varying fields of view (FOVs) is challenging. While existing self-supervised learning methods have shown promising performance, they rely on contrastive learning built upon augmented views of the same image, which inherently limits their generalization in cross-modality scenarios. This motivates research on cross-modality embedding learning using paired multi-modality images. However, existing methods suffer from not only limited training data, but also the requirement of multi-stage iterative training. To address these limitations, we propose UNIVERS, an end-to-end trainable framework that learns UNIversal Voxel-wise Embedding via Randomized Synthesis. Our framework leverages diverse pseudo-data synthesized from organ shape templates, with randomization applied to semantic layout, visual appearance, FOV, and spatial pose. Coupled with cross-domain discriminative voxel-wise contrastive learning, this domain randomization strategy forces the network to disregard domain-specific biases and learn features that are robust against variations in imaging contrast, resolution, and spatial context. Consequently, the trained model enables one-shot landmark localization in multi-modality volume scans with arbitrary FOVs. In this study, the efficacy of the proposed method is demonstrated by applying it to the task of localizing the adrenal glands, which are paired, small, and irregular endocrine organs. In particular, comprehensive experiments are conducted on three challenging adrenal gland datasets. The experimental results demonstrate that UNIVERS consistently outperforms the state-of-the-art (SOTA) one-shot landmark localization approaches, with particularly prominent advantages in cross-modality scenarios. The source code is publicly available at https://github.com/Kendra9999/UNIVERS.
Amodal video segmentation asks for the complete mask of an object in every frame, including the parts hidden behind other objects. The best current methods fine-tune video diffusion models to generate these masks; they need a depth network and tens of denoising steps, and their outputs vary from sample to sample. Self-supervised video world models of the JEPA family are pre-trained on a closely related task, predicting the representations of masked space-time regions from the visible ones, yet their predictor is usually discarded after pre-training. We reuse it by treating occlusion as masking. Tokens not covered by the object's visible mask are dropped, the V-JEPA 2.1 predictor is fine-tuned to fill them in, and a lightweight decoder turns all features into amodal masks. The resulting model, OccJEPA (Occlusion-as-masking JEPA), is deterministic and 32x faster per frame than Diffusion-VAS, the diffusion-based state of the art. It is also more accurate on SAIL-VOS (+3.70 occluded-region mIoU over all occluded frames), on TAO-Amodal (+4.00 AP50 on a held-out test split, +4.32 with SAM2-tracked masks) and on Kubric. Controlled experiments trace most of the gain to the predictor's pre-trained weights rather than its capacity, and the predictor helps at every encoder size tested (87M to 1.85B parameters).
Omni LLMs can naturally locate key audio and visual evidence and integrate information across modalities for audiovisual reasoning. Existing audio-visual reasoning benchmarks that mainly rely on multiple-choice questions usually score models solely based on answer correctness, making it difficult to assess whether a model truly forms a reasoning chain from evidence to the correct answer. In addition, textual or unimodal shortcuts may allow models to answer correctly without analyzing the necessary audio-visual evidence, further weakening the reliability of evaluation benchmarks. To address these issues, we introduce EvaOmni, an evidence-grounded benchmark containing 392 videos and 1,027 purely human-annotated question-answer pairs, covering 15 capabilities and three levels: retrieval and localization, basic understanding, and complex reasoning. We design a series of rigorous annotation procedures in which experts annotate audio and visual evidence and construct video-based distractors; humans review whether questions meet strict inspection criteria, and finally models filter textual and unimodal shortcuts. Beyond high-quality question-answer pairs, we require models to simultaneously return the answer and supporting evidence intervals in order to evaluate the complete reasoning chain. We propose the Coverage Grounded Score (C.G.) to measure how much annotated evidence a model identifies, and the Bottleneck Grounded Score (B.G.) to penalize poor temporal localization in either modality; both are incorporated into answer correctness. Across 7 open-weight and 17 proprietary Omni LLMs, Gemini3.7-Flash achieved the highest overall accuracy (79.4%), C.G. (70.8%), and B.G. (35.1%). At Level 3, it reached 79.1% accuracy, but its B.G. was only 12.3%. These results show that even the most advanced models still have room for development in grounded audio-visual reasoning.
Negation remains a persistent challenge for language models across NLP tasks and modalities. We show that this weakness can undermine LLM safety guardrails through a structural vulnerability we term polarity blindness: the failure to recognize harmful intent after polarity-altering linguistic transformations. We study two forms of polarity inversion: grammatical negation and antonym replacement, and use synonym replacement as a polarity-preserving control. We further test whether an adversarial textual wrapper amplifies these failures. Using a 260-prompt benchmark across 10 state-of-the-art LLMs and 3 languages (English, Spanish, Czech), stratified by harmfulness potential and evaluated on a validated 0-5 harm scale, we find that polarity blindness affects most evaluated models, with mean harm scores reaching 4.12 for the most vulnerable model. The effect persists across typologically differing negation systems, suggesting that polarity blindness reflects a broader limitation in semantic robustness rather than a language- or grammar-specific failure.