arXiv:2605.23543v1 Announce Type: new
Abstract: The Java Stream API aims at increasing developer productivity thanks to an easy-to-read declarative syntax to express computations. It also simplifies parallel computing, providing a high-level abstraction on top of common parallelization aspects. Unfortunately, there is a lack of benchmarks specifically targeting stream-based applications. Such a lack of benchmarks makes it difficult for researchers and developers of the Java class library to optimize the Stream API. Moreover, in the absence of dedicated benchmarks, it is difficult to analyze the performance of streams to suggest developers how to write efficient code using the API.
In this work we present JEDI, a benchmark suite that targets the Stream API. JEDI is automatically generated by converting SQL benchmarks into Java benchmarks. Our code generator supports targets different implementations (both stream-based and imperative) for the same query. The ultimate goal of our benchmark suite -- and the main contribution of this work -- is to analyze the performance of the different implementations to spot inefficient code structures and better alternatives, suggesting best practices to Java developers. Among the multiple implementations we generate, we focus on different parallelization strategies and explain the most efficient parallelization strategies based on characteristics of the processed data. Finally, the code generation producing imperative code defines of a baseline that can guide researchers and Java implementers to optimize the Stream API.
Science Journals
arXiv:2406.14190v5 Announce Type: replace
Abstract: We present a new extended resolution clause learning (ERCL) algorithm, implemented as part of a conflict-driven clause-learning (CDCL) SAT solver, wherein new variables are dynamically introduced as definitions for {\it Dual Implication Points} (DIPs) in the implication graph constructed by the solver at runtime. DIPs are generalizations of unique implication points and can be informally viewed as a pair of dominator nodes, from the decision variable at the highest decision level to the conflict node, in an implication graph. We perform extensive experimental evaluation to establish the efficacy of our ERCL method, implemented as part of the MapleLCM SAT solver and dubbed xMapleLCM, against several leading solvers including the baseline MapleLCM, as well as CDCL solvers such as Kissat 3.1.1, CryptoMiniSat 5.11, and SBVA+CaDiCaL, the winner of SAT Competition 2023. We show that xMapleLCM outperforms these solvers on Tseitin and XORified formulas. We further compare xMapleLCM with GlucoseER, a system that implements extended resolution in a different way, and provide a detailed comparative analysis of their performance.
arXiv:2601.19575v4 Announce Type: replace
Abstract: Recently, the data protection practices of researchers in human-computer interaction and elsewhere have gained attention. Initial results suggest that researchers struggle with anonymization, partly due to a lack of clear, actionable guidance. In this work, we propose simulating re-identification attacks using the approach of red teaming versus blue teaming: a technique commonly employed in security testing, where one team tries to re-identify data, and the other team tries to prevent it. We discuss our experience applying this method to data collected in a mixed-methods study in human-centered privacy. We present usable materials for researchers to apply red teaming when anonymizing and publishing their studies' data.
arXiv:2602.13985v2 Announce Type: replace
Abstract: Artificial intelligence (AI) has demonstrated strong potential in clinical diagnostics, often achieving accuracy comparable to or exceeding that of human experts. A key challenge, however, is that AI reasoning frequently diverges from structured clinical frameworks, limiting trust, interpretability, and adoption. Critical symptoms, pivotal for rapid and accurate decision-making, may be overlooked by AI models even when predictions are correct. Existing post hoc explanation methods provide limited transparency and lack formal guarantees. To address this, we leverage formal abductive explanations, which offer consistent, guaranteed reasoning over minimal sufficient feature sets. This enables a clear understanding of AI decision-making and allows alignment with clinical reasoning. Our approach preserves predictive accuracy while providing clinically actionable insights, establishing a robust framework for trustworthy AI in medical diagnosis.
arXiv:2601.21500v2 Announce Type: replace
Abstract: In many applications of LLMs, natural language responses often have an underlying structure such as representing discrete labels, numerical values, or graphs. Yet, existing decoding and uncertainty estimation methods operate only in language space and largely disregard structural information. We address this by modeling LLM outputs directly in a task-dependent latent structure. By equipping this structure with a dissimilarity measure, we can compute Bayes-optimal responses. These are not selected from sampled generations but are newly synthesized by combining individual responses in the latent space. Across different tasks, Bayes-optimal responses consistently outperform standard decoding methods like beam search. Moreover, quantifying uncertainty via the induced Bayesian risk captures variations in terms of the latent structure and improves alignment with output quality and correctness. Our decision-theoretic framework is applicable to any problem that admits a latent response structure and enables reliable task-aware LLM predictions.
arXiv:2506.15630v2 Announce Type: replace
Abstract: The $h$-version of the finite-element method ($h$-FEM) applied to the high-frequency Helmholtz equation has been a classic topic in numerical analysis since the 1990s. It is now rigorously understood that (using piecewise polynomials of degree $p$ on a mesh of a maximal width $h$) the conditions "$(hk)^p \rho$ sufficiently small" and "$(hk)^{2p} \rho$ sufficiently small" guarantee, respectively, $k$-uniform quasioptimality (QO) and bounded relative error (BRE), where $\rho$ is the norm of the solution operator with $\rho\sim k$ for non-trapping problems. Empirically, these conditions are observed to be optimal in the context of $h$-FEM with a uniform mesh. This paper demonstrates that QO and BRE can be achieved using certain non-uniform meshes that violate the conditions above on $h$ and involve coarser meshes away from trapping and in the perfectly matched layer (PML). The main theorem details how varying the meshwidth in one region affects errors both in that region and elsewhere. One notable consequence is that, for any scattering problem (trapping or nontrapping), in the PML one only needs $hk$ to be sufficiently small; i.e. there is no pollution in the PML.
The motivating idea for the analysis is that the Helmholtz data-to-solution map behaves differently depending on the locations of both the measurement and data, in particular, on the properties of billiards trajectories (i.e. rays) through these sets. Because of this, it is natural that the approximation requirements for finite-element spaces in a subset should depend on the properties of billiard rays through that set. Inserting this behaviour into the latest duality arguments for the FEM applied to the high-frequency Helmholtz equation allows us to retain detailed information about the influence of $\textit{both}$ the mesh structure $\textit{and}$ the behaviour of the true solution on local errors in FEM.
arXiv:2605.23621v1 Announce Type: new
Abstract: The Electron-ion collider in China (EicC) is a proposed future electron-ion collider designed to achieve a high luminosity, with a center-of-mass energy ranging from 15 to 20 GeV. Excellent particle identification (PID) with extensive momentum coverage is essential for investigating exclusive and semi-inclusive processes, as well as enabling precise 3D imaging of the nucleon structure in the EicC experiment. To meet its PID requirement, the EicC Collaboration has proposed the conceptual design of various Cherenkov detectors, including the DIRC in the barrel region and the RICH in the endcap region. It also involving the TOF detector for PID in the low momentum region. The GEANT4 simulation, which integrate advanced optical transmission models and image reconstruction algorithms, have been conducted to study and optimize the performance of these detectors.
arXiv:2601.21513v2 Announce Type: replace
Abstract: In distributed applications, such as energy demand forecasting at the substation level or federated learning, a large number of related tasks must be learned by different models, while the exact task relationships are unknown. We propose the novel Cascaded Transfer Learning (CTL) paradigm in which model parameters cascade hierarchically through tasks organized as a rooted tree, respecting a global training budget. Starting from a source task, the tree specifies the order in which tasks are learned and refined, with the budget allocated along its branches. We design cascade mechanisms based on spanning trees that connect all tasks by minimizing an objective combining pairwise task distances and the available training budget, which yield geometry-aware and depth-bounded transfer graphs. We theoretically characterize how transfer errors accumulate and attenuate along cascade paths: errors introduced at any upstream node are contracted by every downstream refinement, and balanced tree topologies bound this accumulation. Experiments on synthetic and real many-task settings, time-series forecasting and image classification, show that CTL enables more accurate and cost-effective adaptation across large task collections than alternative approaches, with the largest gains at the tightest budgets.
arXiv:2601.21692v2 Announce Type: replace
Abstract: Fine-Tuning-as-a-Service (FTaaS) facilitates the customization of Multimodal Large Language Models (MLLMs) but introduces critical backdoor risks via poisoned data. Existing defenses either rely on supervised signals or fail to generalize across diverse trigger types and modalities. In this work, we uncover a universal backdoor fingerprint-attention allocation divergence-where poisoned samples disrupt the balanced attention distribution across three functional components: system instructions, vision inputs, and user textual queries, regardless of trigger morphology. Motivated by this insight, we propose Tri-Component Attention Profiling (TCAP), an unsupervised defense framework to filter backdoor samples. TCAP decomposes cross-modal attention maps into the three components, identifies trigger-responsive attention heads via Gaussian Mixture Model (GMM) statistical profiling, and isolates poisoned samples through EM-based vote aggregation. Extensive experiments across diverse MLLM architectures and attack methods demonstrate that TCAP achieves consistently strong performance, establishing it as a robust and practical backdoor defense in MLLMs.
arXiv:2601.21703v2 Announce Type: replace
Abstract: Fewest-switches surface hopping (FSSH) is the most popular method for simulating photochemical processes of molecular systems. Recently, we have constructed long short-term memory (LSTM) networks as a propagator for electronic subsystems in FSSH dynamics simulations. The collective results on Tully's three models have been reproduced satisfactorily. In the present work, we develop an extended LSTM-FSSH framework to simulate realistic photochemical reactions. The input features of LSTM as well as the training procedure are redesigned to represent high-dimensional nuclear degrees of freedom in an effective way. Equivariant neural networks are integrated with LSTM to build adiabatic potential energy surfaces in ground and excited states. Photoisomerizations of $\mathrm{CH_2NH}$ and azobenzene are simulated, showing that our new proposed LSTM-FSSH method can produce excited-state lifetimes and product yields accurately in comparison with conventional FSSH simulations as reference. Only 10 reference trajectories are required for training LSTM networks, and then a trajectory ensemble can be generated with very efficient LSTM-FSSH dynamics simulations to obtain collective results.
arXiv:2604.17134v2 Announce Type: replace
Abstract: We present RoIt-XMASA, a multilingual dataset that extends the Cross-lingual Multi-domain Amazon Sentiment Analysis to Italian and Romanian, comprising 36,000 labeled reviews across three domains (books, movies, and music) and 202,141 unlabeled samples. To address cross-lingual and cross-domain challenges, we propose a multi-target adversarial training framework that employs loss reversal with meta-learned coefficients to dynamically balance sentiment discrimination with domain and language invariance. XLM-R achieves an F1-score of 66.23% with our approach, outperforming the baseline by 4.64%. Few-shot evaluation shows that Llama-3.1-8B achieves 58.43% F1-score, revealing a meaningful trade-off between the efficiency of prompting-based approaches and the higher performance of task-specific fine-tuning.
arXiv:2605.23382v1 Announce Type: new
Abstract: Agentic reinforcement learning (Agentic RL) has achieved strong progress in tasks with clear success signals. However, many real-world agent applications require user-conditioned behavior: the same query may call for different planning strategies and tool-use decisions across users. This setting raises key challenges: generic rewards cannot capture heterogeneous user preferences, observed behaviors are entangled with conformity effects, and flat memories cannot support personalized skill retrieval. To this end, we propose a unified personalized Agentic RL framework that embeds personalization into training-time optimization. At its core is \emph{Personalized Anchor Reward-Decoupled Policy Optimization} (\textbf{PARPO}), which decouples generic task-quality rewards from personalized preference rewards and uses user-specific anchors to stabilize learning under heterogeneous reward scales. We further introduce a two-stage preference-disentangled reward model and \emph{Preference-Aligned Skill Evolution Graph Memory} (\textbf{PSGM}) for personalized supervision and preference-aligned skill retrieval. Together, they form a closed loop of preference identification, policy optimization, and structured skill accumulation. Experiments on ETAPP, ETAPP-Hard, and SJAgent show that our framework consistently outperforms strong memory and RL baselines. Code and data are included in the supplementary materials.
arXiv:2310.05999v3 Announce Type: replace
Abstract: Integrated hydrogen-enriched compressed natural gas (HCNG) and active distribution network (ADN) is providing efficient and sustainable flexibility for consuming renewable energies. Yet, cross-sector privacy and uncertain high-renewable scenarios block stable coordination. They also worsen decision performance and convergence. To conquer the barrier, a neural enhanced time-adaptive robust Nash bargaining strategy is proposed.In the first stage, to clear energy trading between ADN and gas distribution network (GDN) and promote its sustainability, a privacy preserved Nash Bargaining based on the alternating direction method of multipliers (ADMM) is applied. The next robust dispatch stage explores the worst renewable scenarios and derisks ADNs profit collapse from uncertainties. The convergence of the entire energy trading scheme is theoretically proved. As such, sustainable returns from the participation of solid oxide fuel cell (SOFC) and HCNG are facilitated. Finally, a time complexity and social welfare co-driven neural-enhanced time-adaptive strategy is proposed. The strategy assesses the influence of time resolution on social benefits and solving time in multi-energy trading. Based on the assessment, a neural network surrogate model is trained to accelerate the trading process in a close looped manner. Numerical assessment reveals that, the proposed strategy reaps a stable social welfare of nearly 1.6% to total cost, and benefit-steady situations for both ADN and GDN, even in the worst renewable scenarios. Moreover, it reduces runtime to 102.47s, improving computational efficiency by over 69.86% versus the fixed time-scale baseline, almost without sacrifice in economy.
arXiv:2605.20201v2 Announce Type: replace
Abstract: Recent large language models support inputs of up to 10 million tokens, yet they perform poorly on long-context tasks that require complex reasoning. Such tasks can be solved using only a subset of the input -- a proxy context -- rather than the full sequence. Despite sharing the same underlying reasoning process, models exhibit a significant performance disparity between proxy and full contexts. To improve long-context reasoning, we propose ProxyCoT, a novel training framework that transfers reasoning capabilities from short proxy contexts to full long contexts. Specifically, we first obtain high-quality chain-of-thought reasoning traces on proxy contexts through reinforcement learning or distillation from a larger teacher model, and then ground the generated traces in full long contexts with supervised fine-tuning. Experiments across different datasets demonstrate that ProxyCoT consistently outperforms strong baselines with reduced computational overhead. Furthermore, models trained with ProxyCoT generalize their long-context reasoning capabilities to out-of-domain tasks.
arXiv:2605.23632v1 Announce Type: new
Abstract: We introduce CopFITi, a copula model for probabilistic forecasting of irregular multivariate time series (IMTS). Our model combines the expressivity of normalizing flows for univariate marginals with the consistency and flexibility of a Gaussian Mixture Copula for the joint dependency structure. Our experiments show that copula-based approaches, which decouple the marginals from the joint, yield better marginal models than architectures that directly fit the full joint. With CopFITi, we propose the first IMTS copula that is marginalization-consistent by construction and establish a new state of the art in joint IMTS density modeling.
arXiv:2412.14642v4 Announce Type: replace
Abstract: Recently, Large Language Models (LLMs) have demonstrated great potential in natural language-driven molecule discovery. However, existing datasets and benchmarks for molecule-text alignment are predominantly built on one-to-one mappings, measuring LLMs' ability to retrieve a single, pre-defined answer, rather than their creative potential to generate diverse, yet equally valid, molecular candidates. To address this critical gap, we propose Speak-to-Structure (S^2-Bench), the first benchmark to evaluate LLMs in open-domain natural language-driven molecule generation. S^2-Bench is specifically designed for one-to-many relationships, challenging LLMs to exhibit genuine molecular understanding and open-ended generation capabilities. Our benchmark includes three key tasks: molecule editing (MolEdit), molecule optimization (MolOpt), and customized molecule generation (MolCustom), each probing a different aspect of molecule discovery. We also introduce OpenMolIns, a large-scale instruction tuning dataset that enables Llama3.1-8B to surpass the most powerful LLMs like GPT-4o and Claude-3.5 on S^2-Bench. Our comprehensive evaluation of 31 LLMs shifts the focus from simple pattern recall to realistic molecular design, paving the way for more capable LLMs in natural language-driven molecule discovery. Our codes and datasets are fully accessible through the Github Repository: https://github.com/phenixace/S2-TOMG-Bench and Huggingface Datasets: https://huggingface.co/datasets/phenixace/S2-TOMG-Bench.
arXiv:2605.23555v1 Announce Type: new
Abstract: This paper addresses the challenge of reconstructing photorealistic and animatable 3D human avatars from monocular videos. While existing methods rely on combining per-subject optimization with generic human priors, they often fail to capture fine-grained details when training frames are limited. To mitigate this data scarcity, we propose TrioMan, a systematic tri-module framework for augmented 3D avatar learning. Our approach comprises three synergistic components. The Generator creates diverse unseen samples by imposing Gaussian perturbations on pose and camera. The Refiner improves the quality of generated data through one-step diffusion guided by texture and geometry cues. The Examiner selects subject-consistent samples using a dual-branch attention-based similarity evaluation. Experiments on the X-Humans and NeuMan benchmarks show that TrioMan outperforms state-of-the-art methods.
arXiv:2605.23821v1 Announce Type: new
Abstract: We propose a distributional theory of how hypernymy -- the ``is-a'' relation between general and specific concepts -- is encoded geometrically in language representations. Starting from the empirically verified assumption that words closer on the WordNet hypernym graph co-occur more often, we characterize theoretically the spectrum of the resulting embedding Gram matrix of word2vec embeddings. Under mild positivity and decay conditions on the co-occurrence kernel, we prove that the leading eigenvectors first separate broad taxonomic branches and then progressively finer sub-branches, producing a \emph{hierarchical splitting geometry} with a coarse-to-fine spectral organization that mirrors the tree. We confirm these predictions in word2vec embeddings across many sampled WordNet subtrees, and show that the same signature extends strikingly well to Gemma 2B unembeddings. Our results indicate that hierarchical concept geometry in LLMs need not reflect a hierarchy-specific functional mechanism, but emerges from the spectral structure of pairwise word statistics.
arXiv:2605.23254v1 Announce Type: new
Abstract: Learning from real-world data is frequently hindered by the compound challenge of long-tailed class distributions and noisy annotations. Existing methods partially address these issues but typically ignore the non-uniform impact of label noise across classes, resulting in ineffective correction for tail classes and over-regularization for head classes. To address this issue, we propose Class-Adaptive Rectification with Experts (CARE), a parameter-efficient framework that leverages three complementary supervision sources from vision-language models (VLM): observed noisy labels, VLM text embeddings, and visual features. CARE introduces a class-adaptive expert consensus mechanism that enforces stricter agreement for tail classes and more permissive agreement for head classes based on class frequency. By aggregating high-confidence predictions across these sources, CARE filters unreliable signals and recalibrates class distributions, yielding more reliable rectification under long-tailed distributions. Extensive experiments on both synthetic and real-world benchmarks demonstrate that CARE consistently outperforms state-of-the-art methods, achieving up to 3.0\% performance gains. The source code is available at https://github.com/qwq123-study/CARE.
arXiv:2605.23860v1 Announce Type: new
Abstract: In this work, a hybrid simulation framework combining TCAD and Allpi$\text{x}^2$ is presented to investigate the sensor properties of MALTA2, a depleted monolithic active pixel sensor designed for future tracking. The study starts from 3D modeling and transient simulations in TCAD, with generic doping profiles and simple well structures. The resulting doping profiles and electric field are extracted and fed into Allpi$\text{x}^2$ for high-statistics Monte Carlo simulations in both DUT-only and full-telescope mode.
Simulations reveal a strong dependence of sensor performance, specifically the detection efficiency and cluster size, on the doping concentration of the N-type blanket at the sensor surface. The doping concentration is then optimized by comparing simulations with measurement data. The active depth of the depleted region of the MALTA2 sensor is estimated in both simulations and measurements using a grazing angle method, in which the sensor is positioned at various inclinations relative to the beam, covering angles from 0 to 60 degrees. Excellent agreement on active depth is obtained with the optimal doping concentration, showing a deviation of 2\% from the measured value at a threshold of 450\,$\text{e}^-$. Consequently, the framework offers a generic toolkit for sensor studies without requiring proprietary information.
arXiv:2605.23861v1 Announce Type: new
Abstract: Causal generative modeling is essential for developing reliable and transparent AI systems capable of counterfactual reasoning. While existing approaches focus on integrating causal constraints during the training of generative models, they often lack a unified framework to leverage the zero-shot reasoning capabilities of pretrained foundation models. We introduce FM-CGM, a modular framework for end-to-end visual causal reasoning using pretrained foundation models. FM-CGM formalizes the causal pipeline through three core components: a concept extractor, a concept manipulator, and a counterfactual generator. By leveraging a large reasoning model for causal inference and a text-to-image diffusion model for generation, our approach enables zero-shot causal discovery, intervention, and counterfactual generation. We then develop Causal Semantic Guidance (CSG), a cross-attention-based mechanism that ensures semantic interventions propagate to descendant concepts while preserving invariant regions. We empirically show that our approach can identify plausible causal structures and is suitable for faithful counterfactual image generation.
arXiv:2605.23843v1 Announce Type: new
Abstract: The commitment-based AKE model provides a formal security framework for key exchange protocols that avoid long-term cryptographic material, achieving authentication through a final out-of-band verification of session-derived values. Within this model, secure KA-based and KEM-based protocols were previously constructed via a commitment-based MT compiler, yielding optimized 4-pass protocols. In this work, we show that 3-pass protocols secure under this model exist for both primitives. These protocols are constructed ad hoc, following the core ideas of the commitment-based MT authenticator, and their SK security in the unauthenticated model is proved using the same game-based techniques, achieving bounds of the same form as those previously achieved. The resulting protocols provide one-way authentication in three message exchanges.
arXiv:2605.23867v1 Announce Type: new
Abstract: Large language models (LLMs) have the potential to aid and improve human decision-making in classification tasks, not only by providing fairly accurate predictions, but also in their ability to generate cogent narrative explanations of those predictions. Prior work has demonstrated that people generally find AI narrative explanations to be understandable, trustworthy, and convincing for changing beliefs and opinions; however, less is known about the impact of narrative explanations on objective human decision-making performance. Here we conduct a large-scale human behavioral experiment to evaluate decision-making performance with LLM-generated narrative explanations of varying persuasiveness. We found the degree of persuasiveness, or lack thereof, for LLM-based explanations did not meaningfully impact decision accuracy over a simple AI prediction alone, in agreement with typical results with explainable AI based on feature importance. We found evidence that narratives increased reliance on AI, but both when the AI prediction was correct and incorrect. Exploratory analyses also indicated that the more persuasive narratives may have had a detrimental effect on decision response times and the ability to discriminate between a correct and incorrect AI prediction. Overall, this work indicates that including narrative explanations with AI predictions may involve tradeoffs for decision-making performance, and more work is needed to determine how and when narrative explanations impact human decision-making.
arXiv:2605.23671v1 Announce Type: cross
Abstract: Real-time hierarchical energy-sharing markets are promising to coordinate large numbers of prosumers. Still, most existing clearing methods rely on linearized or DC power-flow models and do not explicitly handle reactive power or voltage-security constraints. With AC network constraints, the problem becomes a large-scale bilevel Mathematical Program with Equilibrium Constraints (MPEC) that is difficult to solve in real time. This paper develops a non-iterative clearing algorithm for two-layer energy-sharing markets with voltage constraints. We first derive an efficient best-response function for each lower-layer energy-sharing market and reduce the equilibrium search to one dimension by exploiting the pricing-coupling structure. We then embed this function into the upper-layer network-constrained problem and reformulate the bilevel MPEC as a single-level mixed-integer second-order cone program (MISOCP), which is computationally tractable. Case studies on the IEEE 123-bus system with 12,300 prosumers show that the proposed method preserves nodal voltages within prescribed limits and delivers solutions with maximum errors below 0.01\% in 0.829 s.
arXiv:2605.23590v1 Announce Type: new
Abstract: ReAct-style agents for search-intensive, multi-step reasoning tasks rely largely on their own internal judgment to decide what evidence to seek, which reasoning or action step to take next, and when to stop, often producing shallow, redundant, or poorly targeted trajectories. Prior work has explored rubrics as external quality signals, but existing uses are mostly evaluative rather than action-guiding: rubrics typically serve as training-time rewards or post-hoc evaluators of completed outputs, and in deep-research settings they are often coarse-grained and report-level rather than step-level. We introduce Co-ReAct, a rubric-guided action-selection framework that uses rubrics as step-level guidance during inference. At each decision step, Co-ReAct injects a rubric into the agent's context to guide the next Reason-or-Act decision, specifying what the agent should target in evidence seeking, search, reasoning, or self-evaluation. To make this guidance reliable, we train a dedicated rubric generator with GRPO. Unlike prior pairwise or binary preference formulations, our objective optimizes a list-wise Spearman rank-correlation reward against multi-judge expert consensus rankings, encouraging rubrics that are discriminative rather than merely plausible. On DeepResearchBench and SQA-CS-V2, Co-ReAct consistently improves over ReAct and representative test-time compute baselines across search agents built on both 8B/14B open-source and frontier closed-source base models. The trained rubric generator can also serve as a drop-in component that improves these baselines without changing their underlying decision mechanisms. Our code is publicly available at https://github.com/ZBWpro/Co-ReAct.