Retrieve-then-generate pipelines are commonly used to produce deep-research answers for open-ended questions, but retrieval alone is insufficient: LLMs must organize noisy and fragmented evidence into comprehensive, well-cited answers. We refer to this process as evidence synthesis. However, direct generation often underuses evidence, misaligns citations, and collapses diverse information into shallow summaries, exposing an evidence synthesis gap between retrieval and generation. Thus, we propose DeepWeaver, a novel framework that weaves noisy retrieved evidence into comprehensive answers by maintaining Thought Block Chains (TBCs), a structured representation that groups claims, salient information, keywords, and supporting evidence. DeepWeaver uses subordinate TBCs to inspect residual evidence, commit TBC revisions, and discover new claims before final generation. We evaluate DeepWeaver on open-ended QA over both knowledge bases and the web, and introduce LoQA, a high-density benchmark for evidence synthesis. Across multiple LLMs, DeepWeaver improves content sufficiency, citation grounding, and detail preservation on LoQA, while achieving deeper insights and higher citation quality on DeepResearch Bench. These results show that evidence weaving is an effective mechanism for bridging retrieval and generation in open-ended QA. Our code is available at https://github.com/KlozeWang/DeepWeaver.
Digital subtraction angiography (DSA) is the reference standard for leptomeningeal collateral (LMC) assessment, providing critical prognostic insights to guide secondary treatment strategies, neurorehabilitation planning, and retrospective stroke research. However, clinical LMC grading via the ASITN/SIR scale relies on manual, highly variable visual inspection. We introduce X-LMC, a spatiotemporal framework for automated collateral scoring from time-resolved biplane DSA. The proposed architecture encodes spatial frame representations through a DINOv2 backbone, fuses orthogonal projections via a token-level cross-view attention module, and models representations of contrast bolus dynamics using a recurrent network architecture. We evaluate our framework on a multicenter dataset of 134 patients with M1-segment occlusions. In a 5-fold cross-validation setting, X-LMC yields higher point estimates than static architectures and spatiotemporal baselines adapted from related angiographic tasks, achieving a Quadratic Weighted Kappa (QWK) of 0.398 (vs. 0.322) and a dichotomized macro-F1 score of 0.711 (vs. 0.663) against the best-performing baseline. X-LMC performance also aligns with the observed clinical inter-rater agreement (QWK: 0.314). As the first DSA study attempting to automate LMC scoring, we demonstrate that multi-view temporal deep learning can capture collateral-specific contrast kinetics. Ultimately, these benchmarks delineate the clinical ambiguities and achievable performance boundaries of automated ASITN/SIR grading, establishing a reproducible foundation for objective hemodynamic phenotyping in stroke cohorts. Code is available at https://github.com/maedehafezi/X-LMC.
Museum and archival datasets do not mirror historical artistic production, but materialize the contingent histories of collecting, preservation, cataloging, and digitization. This has direct consequences for interpreting pretrained image representations: they may appear to encode historical time while actually encoding the institutional conditions under which objects become visible as data. We describe this phenomenon as temporal entanglement and investigate it by formulating artwork dating as an uncertainty-aware regression task over frozen image embeddings. We evaluate several pretrained vision models on a temporally controlled Wikidata corpus of artworks. Our results show that these models contain usable temporal information, with Vision-Language Models (VLMs) outperforming purely visual self-supervised baselines. However, a qualitative analysis indicates that this temporal knowledge is shaped by various biases.
Bioassay activity prediction is often data-limited because drug-discovery datasets rely on time-consuming and expensive wet-lab experiments for data generation and evaluation. This challenge has inspired recent research into molecular foundation models (MFMs), which aim to encode general-purpose chemical knowledge into molecular representations that generalize well in data-constrained scenarios. This paper presents Monroe, a new MFM with several innovations over the existing state of the art: increased scale allowing pre-training on over 81 million molecules from the PM6 quantum chemistry dataset; improved graph representation of stereochemistry; improved training losses including conformer denoising and embedding decorrelation; improved multi-task learning; and the use of a prior-data-fitted model (TabPFN) for downstream in-context prediction. Our evaluations use a principled pairwise comparison framework that measures statistically significant performance differences. Across established Polaris benchmarks, Monroe matches or exceeds existing MFMs, while on activity cliff benchmarks, designed to assess utility for molecular discovery, it achieves significant improvements over prior methods. Finally, ablation and transfer experiments show that PFN-based downstream predictors also substantially improve two leading existing models, MiniMol and CheMeleon, yielding new state-of-the-art variants we call MiniMol_PFN and CheMeleon_PFN, suggesting that our downstream adaptation strategy generalizes beyond Monroe. Source code is at github.com/blazejba/monroe.
Semantic segmentation demands a careful balance between accuracy, efficiency, and scalability, which remains difficult to achieve for high-resolution imagery. Convolutional networks effectively model local patterns but struggle with long-range dependencies, whereas Vision Transformers capture global context at a high computational cost. While recent work largely focuses on encoder design, the bottleneck stage, central to contextual aggregation and information flow, has been relatively overlooked. We propose SiConMo, a lightweight yet effective framework, implemented in two variants: an RGB-only model (SiConMo) and a GME-enhanced variant (SiConMo$_\dagger$). We show that simplicity arises from a key design principle: at very low computational budgets, the bottleneck is the most efficient stage to integrate local and global context. SiConMo integrates three complementary components: a Token Pyramid Extraction Module for hierarchical multi-scale representation, a Transformer-Branched Depthwise Convolution block for bottleneck-aware context modeling, and a Feature Merging Module that preserves spatial structure while enhancing semantic consistency. Extensive experiments on ADE20K, PASCAL Context, Cityscapes, and COCO-Stuff demonstrate that SiConMo achieves a state-of-the-art accuracy-efficiency trade-off among lightweight semantic segmentation models, highlighting simplicity as a powerful design principle.
Continual Learning (CL) enables deep learning models to iteratively learn from a stream of data without forgetting prior knowledge. Existing adversarial research on CL primarily aims to re-enable catastrophic forgetting, attacking stability and reducing availability. We identify a novel security flaw: data manipulated by an attacker can reduce the learnability of current or upcoming iterations. We term such manipulations learning blockers, as they attack the plasticity of CL algorithms. They are particularly harmful because they are difficult to detect during training of the current iteration, since they can target iterations whose data the model has not yet encountered. When learning blockers additionally induce catastrophic forgetting, the resulting overall degradation is what we call catastrophic learning. We formalize this scenario, define a threat model and propose six attack strategies: Label-Exchange, Tensor-Exchange, Attraction-Coincident, Attraction-Preceding, Repulsion-Coincident, and Repulsion-Preceding. The Attraction variants minimize the loss between the poisoned and the victim iteration label, pulling their representations together in feature space; the Repulsion variants maximize this loss, pushing them apart so stability mechanisms resist the required parameter shift. In the Coincident variants, the poisoned and the victim iteration coincide, using a clean reference iteration only as a label source; in the Preceding variants, the poisoned iteration precedes the victim, leaving it unlearnable due to distorted representations. We evaluate on MNIST and Split-CIFAR10 against three CL strategies - DER, ER-ACE, and iCaRL - across more than 4,480 simulations. Our results demonstrate a strong vulnerability: an adversary can selectively impede plasticity to hinder the acquisition of new knowledge, while promoting loss of prior knowledge, inducing a catastrophic learning scenario.
Historical newspapers are an abundant record of public life, but their dense, irregular and sometimes noisy layouts make computational access to these materials both challenging and limited. We present the Institutional Newspapers Pipeline, a modular system we jointly designed with Boston Public Library to extract high-quality, structured datasets from historical newspaper scans. It was architected so that each step remains interpretable and customizable, and so that the pipeline as a whole remains computationally frugal enough to run on workstation-level hardware. The pipeline runs each scan through a multi-step process: it segments scans into individual type-agnostic crops and performs OCR on each resulting segment before then performing text analysis, type classification, reading order detection, named entities recognition, subject classification, language detection, and pre-computed embeddings generation on every crop. We ran this pipeline against a portion of Boston Public Library's holdings and released the results as an open dataset. The optical character recognition (OCR) output represents 16.3 billion o200k_base tokens across 83.1 million individual crops, extracted from 1,473,635 public domain newspaper scans published between 1795 and 1930. This report describes our methods for each processing step, the small models we trained, as well as the evaluation results and dataset-scale measurements we collected in the process. It accompanies the release of the pipeline, models, and dataset. We position this work as a substantial step towards unlocking high-quality data from tens of millions of newspaper scans.
Weighted programs extend guarded commands with trace weights drawn from a semiring, or more generally a monoid-module. Varying this algebra gives one programmatic syntax for a variety of quantitative and symbolic models. Weakest-preweighting semantics provides a compositional basis for reasoning about those programs.
We present a deductive verification framework based on a weighted assertion language and an intermediate verification language. Its weight domains are ordered structures with implication and coimplication, which let verification conditions express lower- and upper-bound obligations internally. We prove sound translations of core commands and reusable encodings for various proof rules applying to procedure calls and loops. To facilitate automation, we prove soundness of a quantifier elimination procedure for our assertion language. A prototype in the Caesar verifier checks case studies for probabilistic queueing costs, recursive database provenance with cyclic dependencies, clearance bounds for networks of arbitrary size, and formal-language reasoning about lock-freedom of a compare-and-swap counter.
We consider the problem of classification of skin tone using photoplethysmography (PPG) signals with labels of the ordinal six-class Fitzpatrick skin tones. A typical accuracy for this task is a poor 40-55 %. However, the labels are subjectively determined by comparing the skin with a colour chart, and hence contain widespread small-scale inaccuracies. By working with a "fuzzy accuracy", which deems a prediction of skin tone class to be correct if its difference from the labelled class is not greater than one, much higher accuracy is obtained which provides more convincing evidence that skin tone can be accurately predicted from PPG signals. Three machine learning approaches were used, namely deep learning or tree-based approaches on raw PPG signals, deep learning on image representations of the signals generated by the Symmetric Projection Attractor Reconstruction (SPAR) method, and machine learning on features extracted from the signals. The first method also employed a fuzzy version of the cross entropy loss function, which gave the best results. Tree-based models on raw signals give accuracies up to 55 % and higher fuzzy accuracies up to 96 %, while deep learning models on the SPAR images obtained lower results of 44 % accuracy and 85 % fuzzy accuracy. The machine learning on PPG features gave similar results to the SPAR method with accuracy of 42 % and fuzzy accuracy of 87 %. We have shown that classification of skin tone using PPG signals is possible with high fuzzy accuracy which implies that our modelling approach enables accurate prediction of skin tone class within at most one class of the observer's choice of class, from which we conclude that PPG signals are affected by skin tone in a discernible way.
Localized image edits can change a photograph's meaning while leaving most of it authentic, so forensic analysis must identify where an edit occurred. We show that patch-level perturbation responses from frozen DINO encoders are themselves localization maps. Training-free Localization of AI-image Edits from patch-token Drift (TRAIL) applies one global Haar perturbation and maps cosine drift between corresponding patch tokens. On 80 source-disjoint CocoGlide test images, TRAIL reaches .903 patch AUROC versus .912 for the mask-supervised Detective SAM; fixed-threshold Dice is .619 versus .709, while an oracle threshold raises TRAIL to .790. Transferred unchanged to Poisson image interpolation, TRAIL reaches .855 AUROC versus .864, showing that the cue persists without a generator. Across sixteen DINO encoders, the best block lies at normalized depth .80-.94. Global context matters: AUROC falls from .903 globally to .857 for local-in-canvas perturbations and .735 for independently encoded crops. Frozen DINO patch tokens therefore contain a strong late-layer localization signal whose visibility depends on the perturbation and preserved context. Code: https://github.com/VishalJ99/trail-image-edit-localization.
Who Can Make the Action Happen? An Authority-Decomposition Framework for High-Risk Automated Systems
High-risk automated systems distribute control across services, credentials, protected components, and lifecycle mechanisms. Labels such as authorized, approved, privileged, or protected therefore do not answer a basic causal question: which actors can actually make a consequential action occur? This paper provides an action-relative method for deriving which trust-domain coalitions are sufficient to cause protected execution, defined as the occurrence of a designated protected state transition. The framework models components, powers, resources, boundaries, and alternative realization structures; includes update, recovery, override, disablement, and alternative invocation; and separates causal control over execution from control over the authoritative account of an operation. It derives inclusion-minimal sufficient coalitions and tests whether claimed execution boundaries remain independent of designated upstream domains. Cross-domain analytical cases illustrate the method. In a split-control, release-intended, open-state, source-bounded Havenlon protocol model, the ordinary witness requires five trust domains, while certificate replacement yields a three-domain inclusion-minimal known requirement set among source-enumerated protocol witnesses; the Linux domain remains insufficient for the complete transition. Deployed global non-bypassability and boundary-bound veto coverage remain unresolved. The framework is a conceptual and analytical tool. It does not certify implementations, establish deployment security, guarantee complete discovery of hidden powers, or define evidence-verification semantics.
Data-driven biology relies on structured knowledge generated by expert biocurators, yet this work remains largely unrecognized in traditional academic assessments. To bridge this gap, we present the updated APICURON platform, a credit-attribution infrastructure that formally acknowledges these scientific contributions. Rather than relying on delayed batch reporting, the system captures curation events as they happen and transforms them into verifiable units of work. This design allows independent resources to define and update their own recognition models while preserving the historical record of each contribution. For researchers, APICURON highlights recent activity alongside lifetime achievements and connects verified activities to persistent academic profiles via ORCID. APICURON has been successfully integrated across biological knowledgebases and data resources, demonstrating its application to diverse workflows. Extending beyond biodata resources, it also supports recognition of non-traditional research artefacts, including training materials and research software, without imposing a rigid definition of contribution.
Historical book collections contain rich visual elements - such as illustrations, photographs, engravings, and decorative art - that are frequently under-explored in large-scale digitization projects. While Optical Character Recognition (OCR) has standardized the extraction of textual content, these visual components offer a layer of nuance and context that remains largely untapped by automated text extraction workflows. This technical report introduces Institutional Books - Visual Elements, an open-source end-to-end pipeline for detecting, classifying, deduplicating, and captioning visual elements from historical book collections. Alongside this pipeline, we release an initial dataset of 22.6 million visual elements extracted from the 983,004 scanned volumes that comprise the Institutional Books: Harvard Library dataset. This work contributes to ongoing, community-wide efforts to enable new use cases for digitized library collections through computational access, from artificial intelligence model training to digital humanities research.
Large language models can improve recommendation quality by reasoning explicitly over user history and candidate items - for example, extracting a user's preferences or explaining why one item fits better than another - rather than mapping history directly to a ranked list. This reasoning, however, is expensive to repeat on every ranking request and, once produced, is typically consumed once and discarded, leaving it neither reusable across future requests nor easy to inspect or correct as user tastes drift. Our insight is that reasoning does not need to be regenerated at every call if it can instead be compressed once into a compact, structured memory that a lightweight model retrieves from. We propose rEDMRec, which distills a teacher LLM's reasoning into four typed, editable experience channels - long-term preference, short-term context, item-perception, and counterfactual hard-negative comparisons - maintained by an LLM memory controller that performs Add/Delete/Modify/Keep operations and refines entries via K-agent debate. A lightweight student LLM then ranks candidates purely by retrieving from this memory, without invoking the teacher again, decoupling online inference cost from reasoning depth. Across ML-1M, Amazon Beauty, and Steam and ten student backbones, rEDMRec improves HR@1 over zero-shot, few-shot, and RAG on every backbone, and over GraphRAG on most backbones, with Impv up to 13.3% vs. the second-best baseline on ML-1M. Channel ablations show that short-term context is the only channel that helps consistently across capacity tiers, whereas long-term, item-perception, and counterfactual contributions are capacity-dependent (and can reverse on the strongest students); debate-based memory optimization lowers bank duplication by 7.4 percentage points while raising downstream HR@1 by up to +0.029 over six optimization epochs.
Collecting robot hand-object interaction data is costly and embodiment-specific, yet abundant human-object videos remain unusable for robot training. We present RoboEdit, a human-to-robot video editing suite that transforms human manipulation videos into action-consistent, physically plausible robot videos with aligned 3D hand states. To enable scalable supervision, we introduce RoboEdit-ADC, an automatic pipeline that reconstructs and retargets 3D interactions from RGB videos across embodiments. This pipeline generates RoboEdit-14M, a large-scale dataset of 174K aligned video pairs (14M frames) spanning seven robot embodiments, diverse scenes, and interaction types. The core editing engine, RoboEdit-Trans, employs cross-embodiment adaptation modules to preserve temporal coherence while adapting appearance and motion. It further integrates a 3D Robot-State Decoder to recover per-frame hand states for structured motion supervision. Experiments show that RoboEdit achieves state-of-the-art editing quality and supports downstream robot control policies in real-world manipulation tasks. Ultimately, the RoboEdit suite unlocks the vast potential of unlabeled human videos, providing scalable, high-fidelity visual and 3D motion supervision for generalizable robot learning.
Clifford circuits play a foundational role in quantum computing, particularly due to their importance in quantum error correction and fault-tolerant logical synthesis. While these circuits can be efficiently simulated and represented as symplectic matrices, standard synthesis methods-such as the Aaronson-Gottesman algorithm-often yield sub-optimal circuits with excessively high gate counts. In this work, we introduce AlphaClifford, a model-based Reinforcement Learning framework powered by Monte Carlo Tree Search, designed to efficiently synthesize Clifford circuits from the fundamental gate set composed of H, S, and CNOT. By modeling the state space through the algebraic properties of the symplectic group, AlphaClifford effectively explores this combinatorial space to minimize overall circuit cost. For unconstrained Clifford optimization, our approach achieves a consistent reduction in both total and two-qubit (CNOT) gate counts compared to state-of-the-art synthesis heuristics, despite operating with a strictly less expressive gate set. Furthermore, we demonstrate the broad applicability of our framework on two additional tasks: hardware-constrained Clifford transpilation, where we outperform existing RL-based compilers, and as a post-synthesis optimization component within a full Clifford+T logical synthesis pipeline. Our results underscore that model-based RL is highly effective at addressing the combinatorial complexities of quantum compilation, offering a scalable pathway to mitigate hardware constraints in both near-term and future fault-tolerant quantum devices.
Single-step retrosynthesis is a central component of computer-aided synthesis planning, yet its intrinsically one-to-many nature is poorly captured by single-answer evaluation and benchmarking protocols. To address this, we introduce Top-K prompting as a robust training and inference paradigm to better capture diverse, plausible reaction predictions. We compile CREED-CCV-2+USPTO-XL, an ultra-large-scale dataset of ~45.6 million verified reactions to train the C3LM (Chemistry Constraint-Consistent Language Model). By integrating fine-tuning with ChemCensor-based and novelty-oriented rewards, our model achieves state-of-the-art performance on the OOD URSA-expert-2026 benchmark. Further analysis of reaction uniqueness shows that LLMs and conventional models explore complementary reaction spaces, motivating ensemble-based retrosynthesis systems. Overall, our results establish Top-K, plausibility-aware training as a practical new direction for robust future LLM-based synthesis planning.
Vision Language Models (VLMs) have recently emerged as a critical component of multimodal AI systems, enabling joint reasoning over visual and textual inputs in real-world and safety-critical applications. Despite their growing deployment, the robustness of VLMs against adversarial threats remains insufficiently explored, particularly in the context of evasion attacks targeting multimodal alignment. In this work, we investigate the vulnerability of VLMs to adversarial perturbations applied to visual inputs and study two attack settings: untargeted attacks, where the goal is to disrupt the model's interpretation of the original image, and targeted attacks, where the adversary aims to force the model to generate a specific semantic description unrelated to the original image. To efficiently generate adversarial examples, we propose a gradient-based attack method that performs optimization exclusively on the vision encoder of the VLM rather than on the entire multimodal architecture. This design significantly reduces the computational cost and resource requirements of the attack while maintaining strong effectiveness. We evaluate our approach on several open-source VLMs, including Qwen2.5-VL, Granite-Vision, FastVLM, and Phi-3.5-Vision, and show that small, human-imperceptible perturbations can substantially alter the textual interpretation produced by the models. Our findings highlight the vulnerability of modern VLMs to adversarial manipulation and emphasize the need for improved robustness and security mechanisms in multimodal AI systems.
Recent Multimodal Large Language Models (MLLMs) are rapidly evolving into unified understanding and generation (UAG) frameworks. However, extending these unified paradigms to the medical domain is hindered by: the absence of comprehensive training and evaluation benchmarks, and the lack of broadly validated unified medical model. To address these gaps, we present a comprehensive foundation for medical UAG. First, we construct MedUAGCorpus, the largest unified medical understanding and generation dataset to date, comprising over 6 million instances across 14 imaging modalities. Second, we introduce MedUAGBench, a systematic benchmark that expands medical generation evaluation to 12 diverse tasks under standardized protocols. Finally, leveraging these resources, we develop MedUAG, an end-to-end trained unified medical model. Extensive experiments demonstrate that MedUAG achieves strong performance across a wide array of understanding and generation tasks, establishing a competitive baseline and paving the way for next-generation medical multimodal systems.
Designing, implementing, and comparing interpretable architectures requires a formal language to represent them. The most common representations fall short in one of two ways. Symbolic equations give no global view of an architecture at a glance. Probabilistic graphical models and flowcharts do not describe actual tensor manipulations, thus hiding key insights and limiting reproducibility. To close this gap, we introduce a graphical notation for designing interpretable AI architectures, adapted from Penrose tensor notation. This graphical notation gives a global view of an architecture and maps one to one onto PyTorch einsum code. We first use this notation to describe architectures that are interpretable by construction, including concept bottlenecks, sparse probes, prototype networks, neural additive models, and mixtures of linear models. We then diagram the key architectural components of Steerling-8B, a frontier interpretable language model. The diagram yields global insights into the architecture (e.g., showing that Steerling is a residual model), a geometric interpretation of each individual operation, and a direct translation into 33 lines of PyTorch code.
Large language model (LLM) based agents have demonstrated remarkable proficiency in automated software issue resolution, yet they often struggle to resolve issues in a specific repository because they lack project-specific knowledge. Existing self-evolving approaches acquire such knowledge from repository history or online repair trajectories, but they either depend on available historical issue-resolution signals or incur substantial per-issue test-time exploration cost. In this paper, we propose SkillForge, a self-distillation framework that proactively acquires project-specific knowledge from the repository itself. Instead of waiting for real issues to expose project-specific knowledge gaps, SkillForge synthesizes project-specific issues by re-implementing test-covered core functionalities of the repository. By resolving these synthetic issues, SkillForge distills reusable project-specific knowledge into entity-grounded skills and associates them with relevant repository entities for future issue resolution. Extensive experiments using both open-source and closed-source models show that SkillForge consistently improves issue resolution performance over strong baselines. These results demonstrate that proactively acquiring project-specific knowledge before solving real issues substantially improves downstream software issue resolution.
Estimating causal effects under network interference typically assumes that the network used for training and the network used for deployment coincide. In practice, an intervention is run on one population while the question of interest concerns a different population, and the two generally differ in topology, node-covariate composition, and spillover pathways. Transporting a causal effect across networks is therefore a data-fusion problem that no existing algorithm solves. We employ a selection diagram, extended to the network setting so that covariate shift and structural network shift enter as separate selectors, and derive from it a transport formula for the direct, spillover, and total effects in the deployment population. Each formula makes explicit which interventional mechanism is assumed invariant and which observational distribution must be reweighted. We then turn the formulas into TranCE (Transported Causal Effects), a doubly-robust algorithm combining an interventional outcome model, a domain density-ratio correction, and cross-fitted inference. Extensive experiments on two semi-synthetic benchmarks derived from real-world social networks and on a fully real weather-insurance field experiment, where the transported effects are checked against held-out randomized estimates, confirm the effectiveness of our approach. Our findings have the potential to improve intervention strategies in networked systems, particularly in social networks and public health.
Test-time scaling (TTS) improves language model outputs by spending additional inference compute - generating multiple candidates, searching over partial sequences, or iteratively refining drafts. These techniques yield large gains on mathematics and code, but have been developed and stress-tested almost exclusively on tasks where verification is straightforward. We conduct the first compute-normalised comparison of five TTS families across five open-ended generation benchmarks spanning medicine, law, finance, general chat, and creative writing - grounded in a unified framework that decomposes the effectiveness of each method's token budget into exploration and exploitation. The answer depends on which side of that decomposition you examine. Scaling exploration works: the best candidate in the pool improves steadily with compute across all settings. What breaks is exploitation - the step that converts a rich candidate pool into a final output. With state-of-the-art generators, reward models correlate at only $ρ_v \approx 0.12$ with true quality, rendering selection near-random regardless of budget. Tree search amplifies this failure through diversity collapse. Refinement helps on one of five benchmarks; its apparent gains elsewhere are confounded. Only synthesis across candidates (Fusion) consistently improves over single-sample baselines, yet still recovers only ~40% of available quality. The candidate pool is not the bottleneck - choosing from it is.
Existing LRM-DoS methods rely heavily on model feedback to synthesize attack queries, requiring either repeated queries to the target model or training a dedicated attack model. These expensive operations severely weaken attack leverage. In this paper, we propose \emph{search amplification}, a novel, model-feedback-free LRM-DoS paradigm. It employs the conflict count derived from an Satisfiability Modulo Theories (SMT) solver as a low-cost external signal to guide the synthesis of inference-heavy Constraint Satisfaction Problem (CSP) instances. Our key observation is that LRMs depend on trial-and-backtracking search when solving CSPs, where higher SMT conflict counts on a given CSP instance positively correlate with more extensive LRM backtracking search and substantially longer output trajectories. Building on this finding, we propose \textsc{SMTrap}, a lightweight, CPU-only framework. Guided by SMT conflict counts, \textsc{SMTrap} generates inference-heavy CSP queries without model queries, attack-model training, or GPU computation. Evaluations across seven frontier models demonstrate the state-of-the-art LRM-DoS capability of \textsc{SMTrap}, producing DoS effects multiple times stronger than existing baselines. To mitigate the threat of \textsc{SMTrap}, we demonstrate a tool-based mitigation that significantly cuts token usage.
Tabular machine learning benchmarks typically summarize performance by averaging scores, ranks, or pairwise wins across datasets. Such aggregates are useful for selecting robust default models, but they can obscure a different question: which models are necessary to attain peak performance on particular datasets? We argue that benchmark evaluation should also consider the data-centric peak performance frontier, defined by the best statistically supported performance achieved on each dataset. From this perspective, a model may be irreplaceable, sufficient, redundant, or fallible depending on where it lies on the frontier relative to other models. Applying this framework to the TabArena benchmark, we find that common aggregation metrics are highly correlated and largely measure consistency and avoiding failures, while being much less aligned with dataset-level irreplaceability. Consequently, models performing decently across datasets without ever being the best choice are rewarded while models with unique dataset-specific strengths appear mediocre under aggregation. Hence, benchmark progress should be measured not only by improvements on aggregation metrics but also by whether new models expand the set of attainable peak performances across datasets.
Showing 526–550 of 2035 papers
« Previous
Page 22 of 82
Next »