arXiv:2606.07181v1 Announce Type: new
Abstract: Single-step retrosynthesis needs both accurate first-ranked suggestions and candidate lists that are rich enough for downstream selection. We study this as a proposal-selection decomposition. Our system, RETROSPECT, combines a single Transformer proposal model, which we call the ChemAlign Transformer, with a LambdaMART reranker over structural, reaction-template, upstream-score, and optional DFT-derived descriptors. The generator is trained with hybrid root-aligned and random SMILES augmentation, Pre-LayerNorm, tied embeddings, exponential moving average weights, and a differentiable atom-balance auxiliary loss. On the full USPTO-50K test set of 5,007 reactions, the generator reaches 55.00% top-1 and 86.18% top-10 exact-match accuracy with 99.86% top-1 validity. On the merged candidate-pool benchmark used for reranking, which contains 5,007 test products and about 111 candidates per product, a LambdaMART model trained on the structural feature set reaches 59.4% top-1 with 0.7171 mean reciprocal rank. Feature ablations show that upstream proposal score and template-frequency statistics provide most of the reranking signal, while DFT and reaction-center DFT features provide smaller and less consistent gains. These results support a modular view of retrosynthesis: stronger single-model proposal and learned candidate selection are complementary, and the proposal model can serve as a drop-in component for ensemble systems such as RetroChimera (Maziarz et al., 2024)
Science Journals
arXiv:2606.07185v1 Announce Type: new
Abstract: Image tokenizers, from 2D grids to recent 1D sequences, typically encode every image with the same fixed number of tokens. Yet visual complexity is highly heterogeneous, so a uniform budget overspends on simple inputs and underserves complex ones. Existing elastic tokenizers expose variable-length reconstructions, but often leave token length as a deployment-time operating point, a search target, or an external prediction rather than an output of the tokenizer itself. In this work, we ask whether a discrete visual tokenizer can budget itself in one pass. Our central finding is that actionable elasticity requires a representation--allocation co-design: prefixes must remain decodable across budgets, and the tokenizer must learn which prefix each image needs. We propose AdaTok, a self-budgeting discrete 1D tokenizer. AdaTok combines Prioritized Representation Learning, which orders tokens with nested tail masking and resolves budget-dependent semantic shift through Multi-Head LoRA decoder heads, with Adaptive Token Allocation, which trains a lightweight deterministic-group GRPO policy over candidate budgets. Dynamic Pareto Weighting balances fidelity and efficiency during policy training without manual trade-off sweeps. On ImageNet-1K, AdaTok-Full reaches rFID 1.31 at 256 tokens, while AdaTok-Adaptive attains rFID 1.50 using only ~118 tokens on average, outperforming discrete 1D baselines at comparable budgets. In autoregressive image generation, the shorter adaptive representation yields ~2.1x throughput over a fixed 256-token decode, suggesting that visual token count can be learned as a content-conditioned output rather than set as a fixed hyperparameter.
arXiv:2604.27011v2 Announce Type: replace
Abstract: AutoML, intended as the process of automating the application of machine learning to real-world problems, is a key step for AI popularisation. Most AutoML frameworks are not accounting for the potential lack of fairness in the training data and in the corresponding predictions. We introduce \textsc{FairMind}, a software prototype aiming to automatise fairness analysis at the dataset level. We achieve that by resorting to the assumptions of the \emph{standard fairness model}, recently proposed by Ple\v{c}ko and Bareinboim. This allows for a sound fairness evaluation in terms of causal effects, based on \emph{counterfactual} queries involving the target, possibly confounders and mediators, and the different values of an input feature we regard as \emph{protected}. After the necessary data preprocessing, the tool implements a closed-form computation of the effects. LLMs are consequently exploited to generate accurate reports on the fairness levels detected in the training dataset. We achieve that in a zero-shot setup and show by examples the expected advantages with respect to a direct analysis performed by the LLM. To favour applications, extensions to ordinal protected variable and continuous targets and novel decomposition results are also discussed.
arXiv:2606.07190v1 Announce Type: new
Abstract: Reasoning prefixes shape the future trajectory of LLM problem solving, yet existing process reward models usually evaluate them through local step correctness. We argue that correctness is a useful but indirect proxy for the effect we ultimately care about: whether a prefix increases the probability of successful completion. We define this effect as prefix gain, the solve-rate improvement induced by conditioning lightweight student model group on a prefix, and use it to train a Prefix Utility Model (PUM) with a simple pairwise ranking objective. PUM learns outcome-grounded prefix utility and can score both complete trajectories and partial reasoning prefixes. Across Best-of-$N$ selection, beam search, and reinforcement learning on mathematical reasoning, PUM provides a strong prefix-level supervision signal, especially when candidate pools are large, search budgets increase, or rule-based rewards are sparse. We release all data, models, and code at https://zhiqix.github.io/pum-project-page.
arXiv:2606.07193v1 Announce Type: new
Abstract: Reinforcement learning (RL) policies enable dynamic legged locomotion but lack mechanisms to avoid violations of safety constraints that are absent during training. Large-scale offline safe learning is impractical for covering all edge cases. Existing safety frameworks either rely on reduced-order models that cannot reason about whole-body behaviors or require conservative recovery controllers that degrade task performance. We propose a predictive safety filter that post-hoc filters the nominal contact locations fed to the RL policy. When a collision is predicted, a sampling-based optimizer asynchronously searches for safer contact sequences using a full-physics model, while a learned value function bootstraps long-horizon returns. Our three algorithmic components (geometric projection of sampled contacts, momentum-augmented updates, and replica-exchange) make the optimization tractable in a discontinuous contact landscape. We validate the filter on a quadruped robot in dense, cluttered environments, both in simulation and in the real world, showing substantial reductions in safety violations with minimal deviation from the nominal input.
arXiv:2606.07196v1 Announce Type: new
Abstract: Classical sparse Type-II Bayesian methods for M/EEG brain imaging support joint estimation of source and noise hyperparameters, but rely on fixed iterative update rules. Although these updates are principled and interpretable, their dynamics cannot be adapted from data. We propose to learn the update mechanism itself while preserving the underlying Bayesian structure by unfolding a classical joint hyperparameter-learning solver into a trainable neural architecture whose layers mirror the original iterations. The resulting framework is initialized to recover the classical solver exactly before training and is enriched through progressively more expressive correction-learning mechanisms, ranging from learnable biases to adaptive MLP and attention-based contextual refinements. In this way, training does not replace Bayesian inference with a black-box predictor, but instead learns structured correction terms while retaining the interpretability and model-based character of the original update dynamics. Structured correction learning therefore aims to improve empirical reconstruction performance without replacing the original model-based inference mechanism. Experimental results show that the learned correction variants improve reconstruction performance and convergence behavior over the baseline unfolded solver while preserving its algorithmic transparency.
arXiv:2606.07201v1 Announce Type: new
Abstract: Two analytical models for sizing the toroidal field coils and central solenoid of a tokamak are developed within the D0FUS system code: a pedagogical thin-cylinder model and a refined thick-cylinder and winding packs model. The refined model shows good agreement with six reference machines and the MADE magnet design code. When the high-field design space is explored for DEMO-class power plants (2 GW of fusion power, Q = 40, tplateau = 2 h), pushing the peak field at the Toroidal Field (TF) coil conductor up to Bmax = 20 T, the radial build emerges as the dominant constraint: in fact, in the baseline wedging/316L configuration, no viable design can be found beyond 20 T, making alternative strategies necessary. The primary levers identified are high-strength steels (e.g CHSN01), alternative mechanical architectures (bucking, plug), and reductions of the effective Central Solenoid (CS) flux demand (for example through auxiliary heating during ramp-up), each carrying an impact of the same order of magnitude on the minimum feasible major radius. Secondary optimisations (conductor shape, radial grading) are shown to provide additional but more modest gains. When all favourable levers are combined (CHSN01, bucking, etc.), compact machines (R0 < 4 m) appear feasible. This suggests that, provided one accepts the associated risks of combining new approaches (CHSN01, bucking, etc.), high-temperature superconductors could unlock the path to compact electricity generating tokamaks.
arXiv:2606.07211v1 Announce Type: new
Abstract: Autonomous robotic systems are being proposed for use in hazardous environments, often to reduce the risks to human workers. In the immediate future, it is likely that human workers will continue to use and direct these autonomous robots, much like other computerised tools but with more sophisticated decision-making. Therefore, one important area on which to focus engineering effort is ensuring that these users trust the system. Recent literature suggests that explainability is closely related to how trustworthy a system is. Like safety and security properties, explainability should be designed into a system, instead of being added afterwards. This paper presents an abstract architecture that supports an autonomous system explaining its behaviour (explainable autonomy), providing a design template for implementing explainable autonomous systems. We present a worked example of how our architecture could be applied in the civil nuclear industry, where both workers and regulators need to trust the system's decision-making capabilities.
arXiv:2605.05220v3 Announce Type: replace
Abstract: Steering intermediate representations has emerged as a powerful strategy for controlling generative models, particularly in post-deployment alignment and safety settings. However, despite its empirical success, it currently lacks a comprehensive theoretical framework. In this paper, we bridge this gap by formalizing the theory of concept steering. First, we establish a link between steering and affine concept erasure, proving that the standard approach for removing unwanted behaviors is a special case of LEACE (a closed-form method for affine erasure). Next, we formulate a principled theoretical framework for concept switching, LEACE-Switch, and characterize the assumptions under which it provides an optimal affine solution. Building on this analysis, we then introduce MidSteer (Minimal Disturbance concept Steering), a more general affine framework for concept manipulation that relaxes these assumptions and enables directed, minimal-disturbance transformations. We demonstrate that MidSteer performs favorably across a range of tasks, modalities, and architectures, including vision diffusion models and large language models.
arXiv:2605.05225v3 Announce Type: replace
Abstract: Mixture-of-Experts Multimodal Large Language Models (MoE MLLMs) suffer from a significant efficiency bottleneck during Expert Parallelism (EP) inference due to the straggler effect. This issue is worsened in the multimodal context, as existing token-count-based load balancing methods fail to address two unique challenges: (1) Information Heterogeneity, where numerous redundant visual tokens are treated equally to semantically critical ones, and (2) Modality Dynamics, where varying visual to text ratios across tasks lead to resource misallocation. To address these challenges, we propose MACS (Modality-Aware Capacity Scaling), a training-free inference framework. Specifically, MACS introduces an Entropy-Weighted Load mechanism to quantify the semantic value of visual tokens, addressing information heterogeneity. Additionally, the Dynamic Modality-Adaptive Capacity mechanism allocates expert resources based on the real-time modal composition of the input. Extensive experiments demonstrate that MACS significantly outperforms existing methods on various multimodal benchmarks, providing a novel and robust solution for the efficient deployment of MoE MLLMs in EP inference.
arXiv:2605.06215v2 Announce Type: replace
Abstract: Multiconfiguration pair-density functional theory (MC-PDFT) provides an efficient and accurate framework for computing electronic energies in strongly correlated molecular systems, with the quality of the on-top functional being a key determinant of its predictive accuracy. Here, we introduce MMCDDB26, a rigorously curated benchmark database comprising 76 datasets and 1,495 reactions. We further propose a constrained, large-language-model-assisted optimization workflow for the development and assessment of MC-PDFT functionals. Using this workflow, we optimized the parameters of the MC23/MC25 functionals on MMCDDB26 to obtain MC26. Compared with earlier functionals of the same class, MC26 improves the accuracy on the training set and achieves a more balanced overall performance. In addition, we developed the hybrid meta-functional COF26. We find that COF26 delivers superior performance for both strongly and weakly correlated systems, and therefore recommend COF26 for future MC-PDFT calculations.
arXiv:2605.08717v2 Announce Type: replace
Abstract: Software engineering agents are increasingly deployed in evaluable engineering environments, yet post-failure recovery remains costly, manual, and ad hoc. Existing systems expose traces or generate follow-up feedback, but they do not convert heterogeneous runtime evidence into grounded, bounded recovery guidance for a subsequent attempt. We present PROBE, a failure-anchored framework for structured recovery in software engineering agents. PROBE organizes failed-run telemetry into structured evidence, structured diagnosis, and bounded recovery guidance through a Telemetry Layer, a Diagnosis Layer, and a Guidance Gate. The Telemetry Layer preserves fine-grained runtime signals, the Diagnosis Layer fuses cross-signal evidence into grounded diagnoses, and the Guidance Gate produces diagnosis-derived guidance only when it is evidence-grounded, actionable, and within the scope of agent-side behavior.
We evaluate PROBE across three settings: repository-level software repair, enterprise workflow recovery, and AIOps service mitigation. On 257 initially unresolved cases, PROBE achieves 65.37% Top-1 diagnosis accuracy and a 21.79% recovery rate, outperforming the strongest non-PROBE baseline by 43.58 and 12.45 percentage points. The results reveal a diagnosis-recovery gap: accurate diagnosis is necessary but insufficient unless translated into bounded guidance that a subsequent attempt can execute and verify. Beyond controlled evaluation, a Microsoft IcM prototype shows that PROBE can attach as a non-intrusive side channel to existing service-diagnosis workflows without changing the agent policy, toolset, or execution budget. These results suggest that telemetry-grounded, failure-anchored recovery can improve post-failure recoverability under realistic engineering constraints.
arXiv:2605.08732v2 Announce Type: replace
Abstract: Modern vision-based world models can represent observations as compact yet expressive latent manifolds, but fast goal-oriented planning in these spaces remains challenging. This raises a central question: when does a learned representation simplify control, rather than merely enabling prediction? We study this question in a pretrained LeWorldModel, whose latent geometry is regularized for smoothness and uniformity. Our key insight is that, under such geometry, planning can be amortized into a latent inverse-dynamics mapping instead of requiring online search. We therefore replace iterative planning with a lightweight Goal-Conditioned Inverse Dynamics Model (GC-IDM) that maps the current latent state, goal latent state, and remaining horizon directly to the next action. Empirically, across four benchmark environments spanning navigation, contact-rich manipulation, and continuous control, our controller matches or exceeds CEM in seven of eight environment-protocol settings while reducing per-decision cost by 100-130x. A broader sweep over test-time planners (CEM, MPPI, iCEM, and gradient-based methods) shows that this result is not specific to a particular optimizer. These findings suggest that much of the structure recovered by test-time planning is already locally encoded in the latent representation. More broadly, our results indicate that sufficiently structured latent spaces can shift part of the planning burden from online optimization to learned inference. Our code is publicly available at https://github.com/hdnndh/Latent-Geometry-Beyond-Search-Amortizing-Planning-in-World-Models .
arXiv:2606.07283v1 Announce Type: new
Abstract: For Human-AI Interaction (HAII) research to move forward, theoretical work linking psychological mechanisms to interface design is needed. Such work should extend rather than replace established HCI and automation research, adapting to the increasing autonomy and agency of AI systems. Building on prior frameworks focused on roles and levels in human interaction with automation, a gap remains from a psychological view: a task-centered, process-oriented account that links mechanisms of action regulation to concrete design and evaluation levers for human-AI coupling, expressed in a unified vocabulary for human and machine. Moreover, existing models may describe how a system is designed (e.g., function allocation in automation) but fall short in showing how this design affects human behavior. We present the Integrated Information Processing (IIP) model, a task-centered, cybernetic model that conceptualizes humans, machines, and their joint activity as coupled control loops. The IIP model uses a unified modeling language for human and artificial agents, making psychological models of action regulation accessible for AI system design. As a core feature, we argue that efficacy within a shared task is characterized by three integration qualities, input adequacy, reference consonance, and output operativity, which critically influence benchmarks of human-centeredness such as transparency and controllability. The model maps interface choices (e.g., XAI techniques) to theory-driven expectations of user behavior, guiding interface design and evaluation. To this end, we present (1) a continuity-preserving theoretical discourse that extends HAII to agency in AI; (2) the IIP model with three information-processing qualities; and (3) applications of the IIP model to exemplary use cases demonstrating implications for interface design.
arXiv:2606.07220v1 Announce Type: new
Abstract: We present a novel adaptive isogeometric method for solving the biharmonic equation over planar multi-patch domains with possibly extraordinary vertices, parametrized by analysis-suitable G^1 multi-patch geometries. The proposed technique relies on the concept of Dual-Primal Isogeometric Tearing and Interconnecting (IETI-DP), which enforces the required C^1-smoothness of the solution across a common edge of two neighboring patches by imposing appropriate continuity conditions by means of Lagrange multipliers. The resulting saddle point problem is solved using a dual-primal formulation, first by a small linear problem for the Lagrange multipliers and then by local, parallelizable linear problems on the individual patches for the coefficients of the numerical solution. While for the local problems on the single patches standard diagonally scaling is used as preconditioner, a suitable preconditioner for the problem of finding the Lagrange multipliers is introduced. To perform adaptive refinement, the solution of the biharmonic equation on the single patches of the multi-patch domain is discretized by employing (truncated) hierarchical B-splines, and an appropriate refinement strategy of the underlying mesh is presented. Finally, the potential of the developed adaptive IETI-DP method for solving the biharmonic equation over planar multi-patch geometries is numerically tested on the basis of several numerical examples. Thereby, the numerical results show on the one hand optimal convergence behavior with respect to adaptive refinement, and on the other hand a good performance of the proposed preconditioner for the linear problem of determining the Lagrange multipliers.
arXiv:2606.07285v1 Announce Type: new
Abstract: We study proportionally fair clustering, where a set of $k$ centers must be chosen from a metric space to represent $n$ agents, and no sufficiently large group of agents should be collectively underrepresented. One of the central notions of fairness in this setting is the $\alpha$-core. The existence of clusterings in the $(1+\sqrt{2})$-core was established by Chen et al. [2019], who also showed instances where the $\alpha$-core is empty for every $\alpha < 2$. Closing this gap has remained an open problem for seven years. We make progress from the lower-bound side by providing an instance whose $\alpha$-core is empty for every $\alpha < 2.1508$. Our techniques rely on establishing connections between variants of the core, namely the Hare core and the Droop core; reducing the search for optimal empty-core instances to a highly structured family of clustering instances; and using a Mixed Integer Linear Program (MILP) to search for optimal lower-bound instances within this reduced space. Using this framework, we also determine tight bounds for Droop quota clustering instances with a small number of possible candidate centers and a single center to be selected. For each number of centers $m \in \{3,4,5,6\}$, we give the exact threshold $\alpha_m^*$ such that an $\alpha_m^*$-core clustering always exists, while for every $\alpha < \alpha_m^*$ there is an instance with $m$ centers whose $\alpha$-core is empty. Although these values were originally found through computer-aided search, we also provide direct proofs that do not rely on MILP certificates.
arXiv:2606.07222v1 Announce Type: new
Abstract: Cell detection in histopathology images strongly depends on surrounding tissue context, where visually similar cells may belong to different classes under different microenvironments. Recent tissue-aware methods incorporate contextual priors, but often rely on static fusion strategies that may propagate noisy information. In this work, we propose DualGate-Net, a prior-aware dual-encoder framework that combines a ConvNeXtV2-based local encoder and a SegFormer-based global encoder through a learnable prior-gated fusion mechanism. The proposed module adaptively regulates the influence of tissue priors across spatial locations, while an auxiliary foreground reconstruction branch preserves high-frequency cellular structures during training. In addition, auxiliary cellness-guided cues are incorporated to further improve localization robustness. Experiments on the OCELOT benchmark demonstrate consistent improvements, achieving macro F1-scores of 0.7722 on the validation set and 0.7345 on the test set, highlighting the effectiveness of adaptive prior integration for robust histopathology cell detection.
arXiv:2605.19611v2 Announce Type: replace
Abstract: Inverse design of metasurfaces under continuous electromagnetic constraints requires generation of geometries that simultaneously satisfy stringent spectral specifications and remain manufacturable. Conventional approaches based on iterative full wave simulations are computationally prohibitive for large design spaces, while existing generative models often suffer from poor conditional controllability and limited fabrication awareness. In this regard, we propose a physics guided condition quality enhanced diffusion framework for the inverse design of metasurface based absorbers. Fabrication-aware constraints are incorporated to ensure practical realizability of the generated designs. The framework introduces a conditioning mechanism for continuous spectral specifications, wherein feature-wise linear modulation propagates the condition across the denoising hierarchy, enabling stable and accurate generation with improved spectral controllability. Further, to embed EM consistency directly into the generative learning process, a pre trained surrogate EM simulator is integrated within the diffusion training pipeline. The proposed framework generated physically realizable metasurface designs for diverse reflection characteristics in the frequency range of 2 to 18 GHz, achieving a very low average spectral mean squared error of 0.0006 and a high band alignment accuracy of 0.958. The framework also addresses the fundamentally non-unique nature of inverse EM design by enabling structured multimodal generation of geometrically distinct yet spectrally consistent metasurface designs for the same target response. The proposed model produces the suitable design in approximately 30 seconds, whereas the conventional approach can take several months under comparable computational resources. The efficiency of the model is also established via experimental measurements.
arXiv:2606.07233v1 Announce Type: new
Abstract: LiDAR-based 3D Multi-Object Tracking (MOT) typically relies solely on geometric information, which is often insufficient to distinguish between targets during prolonged occlusions or in crowded human-populated environments. While integrating RGB-based Re-Identification (ReID) offers a theoretical solution for preserving identity context, existing approaches often rely on computationally expensive parallel detectors that hinder real-time robot responsiveness. This work presents a systematic study of image-based ReID in online 3D MOT, utilizing a lightweight projection-based framework to decouple geometric and appearance modeling for mobile robots. A comprehensive analysis of feature extraction architectures is conducted, employing lightweight CNNs and Vision Transformers, and evaluating various multi-modal data association strategies to balance computational latency with robust tracking. Experiments on the Pedestrian class of the KITTI dataset reveal that naive linear fusion, of appearance and motion costs, degrades performance due to visual noise. Conversely, a cascaded matching strategy successfully recovers occluded tracks without compromising overall precision, effectively preventing identity switches to maintain human-robot interaction continuity. We show that lightweight architectures can offer an optimal trade-off between the low latency required for safe navigation and the discriminative power needed for social awareness.
arXiv:2606.07308v1 Announce Type: new
Abstract: We study off-policy evaluation (OPE) under strategic behavior where decision subjects (or agents) respond to a decision maker's policy by strategically modifying their covariates. Such behavior induces a policy-dependent covariate shift, breaking the standard assumption in existing methods that covariates are exogenous to the policy. Related work addresses this challenge by imposing strong assumptions such as repeated interactions or full knowledge of agents' response behavior, substantially limiting its applicability to OPE. In contrast, we consider a one-shot OPE setting where the decision maker has only partial knowledge of the agents' response behavior. Our key insight is that disclosing local information through post-hoc explanations reveals agents' pre-strategic covariates prior to adaptation, mitigating the information loss induced by strategic behavior. Leveraging this structure, we estimate a statistical model for the agents' responses and construct a doubly robust estimator for policy value. By assuming that the agents' cost sensitivity follows a conditional log-normal distribution, we establish consistency of the proposed estimator and validate our approach empirically. More broadly, our results highlight how interaction design can mitigate information asymmetry by revealing otherwise hidden structure in agents' strategic responses.
arXiv:2604.13174v2 Announce Type: replace-cross
Abstract: Elemental abundance variations in the solar corona, commonly characterised by First Ionisation Potential (FIP) bias, provide crucial diagnostics of chromospheric processes. The ponderomotive force model, which attributes fractionation to Alfv\'en wave propagation, has successfully reproduced the observed fractionation patterns in various solar features. However, existing theoretical implementations rely on a static quiet Sun chromosphere, leaving the influence of chromospheric dynamics largely unexplored. We address this limitation by combining hydrodynamic simulations from HYDRAD with ponderomotive force calculations through FIPpy, a new open-source code. Comparing predictions between an initial VAL-C chromosphere and a heated chromosphere following impulsive nanoflare-like events, we show that the ponderomotive force model remains consistent under dynamic chromospheric conditions, while stronger changes in fractionation behaviour arise from variations in acoustic flux and turbulence. Most significantly, when acoustic wave flux drops below $\sim5\times10^6$ erg cm$^{-2}$ s$^{-1}$, mass-dependent thermal velocities dominate the fractionation process, producing counterintuitive patterns where Fe exceeds Ca in FIP bias, while high-FIP Ar shows fractionation. We demonstrate that any source of chromospheric turbulence will act to suppress fractionation. For flares, our results predict that the increased turbulence will suppress FIP bias, potentially explaining the observed abundance variations during flares. These findings suggest that coronal abundances and composition encode a sensitive balance between dominant mechanisms, determined by the ratio of ponderomotive acceleration to turbulent velocity.
arXiv:2606.07258v1 Announce Type: new
Abstract: Binding prediction models accelerate therapeutic antibody and TCR discovery, but their performance on new datasets is unpredictable, often leading to low discovery rates. Density-ratio methods (PAPE, M-CBPE) provide label-free performance estimation for binary classification, but their assumptions and aggregate-only outputs limit binding prediction on neoepitopes, antigen variants and chemical scaffolds. Here we present CaliPPer (Calibration and Prediction of Performance), a post-hoc framework pairing a multi-chain Sample-to-Domain Distance (S2DD) with distance-aware Bayesian recalibration, operating at three resolutions: generalisability score, aggregate performance prediction, and per-sample confidence. Across ten models, eight architectures and two immune-receptor domains, CaliPPer attains distance--performance correlations $|r|=0.80\text{--}0.92$, predicts AUROC/AP/F1 with mean absolute errors $0.008\text{--}0.070$, and improves AUROC by up to $+0.20$ on unseen epitopes/variants. Applied retrospectively to five published TCR, BCR, MHC--peptide and small-molecule studies, CaliPPer raises true discovery rates in all five (e.g.\ $0/5 \to 3/5$ confirmed neoantigens), providing a triage layer between computational prediction and experimental validation.
arXiv:2605.21347v3 Announce Type: replace
Abstract: Diagnosing failures in LLM agents remains largely manual. Practitioners inspect a small subset of execution traces, form ad-hoc hypotheses, and iterate. This process misses patterns that only emerge across trace populations and does not scale to production corpora where individual traces span tens of thousands of tokens. We formalize the problem of corpus-level trace diagnostics. Given a corpus of execution traces, the goal is to produce grounded natural-language insights that characterize systematic behavioral patterns across trace groups, each linked to supporting evidence. We present the Insights Generator (IG), a multi-agent system that answers diagnostic questions by proposing and testing hypotheses across the trace corpus to produce an evidence-backed insights report. We evaluate IG across qualitative and objective dimensions, spanning rubric-based report assessment and downstream performance improvements achieved by implementing IG insights. Human experts using IG reports improve scaffold performance by 30.4pp over the unmodified baseline scaffold, and coding agents leveraging IG-derived insights show consistent and stable gains. Across benchmarks, IG's scout-investigator architecture produces findings comparable in detection coverage to competing approaches, while domain experts rated IG reports as leading depth and evidence quality.
arXiv:2606.07366v1 Announce Type: new
Abstract: Self-driving simulations typically rely on data collected in a small number of cities or on hand-authored synthetic scenarios. Dashcam videos cover a far broader range of locations and situations, including rare or long-tailed scenarios. They are considered less usable for simulation because it is difficult to recover accurate 4D scenes from monocular in-the-wild videos. Work zones are one such class of long-tailed situations that dashcams capture. We present Dash2Sim, a framework that turns in-the-wild monocular dashcam videos into metric, geo-referenced 4D driving logs compatible with existing simulators, and verifies eachone against an independently maintained map without annotations. We apply Dash2Sim to a large video corpus to create the ROADWork4D benchmark dataset, which spans 4,244 scenes with 2.7M 3D objects across 17 cities. On a verified subset ROADWork4D-CL (2,201 scenes), we study privileged closed-loop planners and find that work zone scenarios are difficult: while rule-based and hybrid planners generalize better than learning-based ones, all fall short, failing to make the lane changes that temporary work zone channels require. Beyond planning, dense depth recovered by Dash2Sim improves novel-view synthesis quality by up to 19% on perceptual metrics, suggesting its potential to provide rich conditioning for closed-loop sensor simulation from monocular videos.
arXiv:2606.07392v1 Announce Type: new
Abstract: Motivated by Large Language Model (LLM) cascading, we propose an online contextual Pandora's Box model for adaptively querying and selecting LLM APIs. In each period, a decision-maker observes a request context and faces a two-phase decision problem. In the query phase, the decision-maker sequentially queries APIs, where each query reveals a generated output and the decision-maker incurs an (output-dependent) cost. In the selection phase, the decision-maker selects one of the generated outputs to deploy and observes only the downstream reward of the deployed output. This output-mediated feedback structure differs from classical online contextual Pandora's Box models, in which opening a box directly reveals its reward.
Rather than estimating the full conditional output and cost distributions of each API, we directly model the reservation index and develop a learning approach for the query phase. Specifically, we impose a parametric structure on the contextual reservation index functions induced by the classical Weitzman's policy. Our policy combines generalized method of moments (GMM) type estimation of these reservation indices with UCB-style confidence bounds for both these indices and the shared output-level reward evaluator. Under regularity conditions, we prove that the resulting policy achieves dimension-dependent $\widetilde O(\sqrt T)$ cumulative regret over a horizon of $T$ periods.