arXiv:2602.08142v2 Announce Type: replace
Abstract: Machine learning applications require fast and reliable per-sample uncertainty estimation. A common approach is to use predictive distributions from Bayesian or approximation methods and additively decompose uncertainty into aleatoric (i.e., data-related) and epistemic (i.e., model-related) components. However, additive decomposition has recently been questioned, with evidence that it breaks down when using finite-ensemble sampling and/or mismatched predictive distributions. This paper introduces Variance-Gated Ensembles (VGE), an intuitive, differentiable framework that injects epistemic sensitivity via a signal-to-noise gate computed from ensemble statistics. VGE provides: (i) a Variance-Gated Margin Uncertainty (VGMU) score that couples decision margins with ensemble predictive variance; and (ii) a Variance-Gated Normalization (VGN) layer that generalizes the variance-gated uncertainty mechanism to training via per-class, learnable normalization of ensemble member probabilities. We derive closed-form vector-Jacobian products enabling end-to-end training through ensemble sample mean and variance. VGE matches or exceeds state-of-the-art information-theoretic baselines while remaining computationally efficient. As a result, VGE provides a practical and scalable approach to epistemic-aware uncertainty estimation in ensemble models.
Science Journals
arXiv:2602.08498v2 Announce Type: replace
Abstract: Large Reasoning Models (LRMs) increasingly rely on reasoning traces with complex internal structures. However, existing work lacks a unified answer to three fundamental questions: (1) what defines high-quality reasoning, (2) how to reliably evaluate long, implicitly structured reasoning traces, and (3) how to use such evaluation signals for reasoning optimization. To address these challenges, we provide a unified perspective. (1) We introduce the ME$^2$ principle to characterize reasoning quality along macro- and micro-level concerning efficiency and effectiveness. (2) Built on this principle, we model reasoning traces as directed acyclic graphs (DAGs) and develop a DAG-based pairwise evaluation method, capturing complex reasoning structures. (3) Based on this method, we construct the TRM-Preference dataset and train a Thinking Reward Model (TRM) to evaluate reasoning quality at scale. Experiments show that thinking rewards serve as an effective optimization signal. At test time, selecting better reasoning leads to better outcomes (up to 19.3\% gain), and during RL training, thinking rewards enhance reasoning and performance (up to 3.9\% gain) across diverse tasks. Code and data are available at https://github.com/Simplified-Reasoning/TRM.
arXiv:2602.09028v2 Announce Type: replace
Abstract: This paper reformulates Fermat's Last Theorem as an embedding problem of information-geometric structures. We reinterpret the Fermat equation as an $n$-th moment constraint, constructing a statistical manifold $\mathcal{M}_n$ of generalized normal distributions via the Maximum Entropy Principle. By Chentsov's Theorem, the natural metric is the Fisher information metric ($L^2$); however, the global structure is governed by the $L^n$ moment constraint. This reveals a discrepancy between the local quadratic metric and the global $L^n$ structure. We axiomatically define an "Information-Geometric Fermat Solution," postulating that the lattice structure must maintain "dual lattice consistency" under the Legendre transform. We prove the non-existence of such structures for $n \ge 3$. Through the Poisson Summation Formula and Hausdorff-Young Inequality, we demonstrate that the Fourier transform induces an alteration of the function family ($L^n \to L^q$, where $1/n + 1/q = 1$), rendering dual lattice consistency analytically impossible. This identifies a geometric obstruction where integer and energy structures are incompatible within a dually flat space. We conclude by discussing the correspondence between this model and elliptic curves.
arXiv:2602.09075v4 Announce Type: replace
Abstract: In-Context Learning (ICL) in transformers acts as an online associative memory and is believed to underpin their high performance on complex sequence processing tasks. However, in gated linear attention models, this memory has a fixed capacity and is prone to interference, especially for long sequences. We propose Palimpsa, a self-attention model that views ICL as a continual learning problem that must address a stability-plasticity dilemma. Palimpsa uses Bayesian metaplasticity, where the plasticity of each attention state is tied to an importance state grounded by a prior distribution that captures accumulated knowledge. We demonstrate that various gated linear attention models emerge as specific architecture choices and posterior approximations, and that Mamba2 is a special case of Palimpsa where forgetting dominates. This theoretical link enables the transformation of any non-metaplastic model into a metaplastic one, significantly expanding its memory capacity. Our experiments show that Palimpsa consistently outperforms baselines on the Multi-Query Associative Recall (MQAR) benchmark and on Commonsense Reasoning tasks.
arXiv:2511.12581v2 Announce Type: replace
Abstract: Static IR drop analysis is a fundamental and critical task in the field of chip design. Nevertheless, this process can be quite time-consuming, potentially requiring several hours. Moreover, addressing IR drop violations frequently demands iterative analysis, thereby causing the computational burden. Therefore, fast and accurate IR drop prediction is vital for reducing the overall time invested in chip design. In this paper, we firstly propose a novel multimodal approach that efficiently processes SPICE files through large-scale netlist transformer (LNT). Our key innovation is representing and processing netlist topology as 3D point cloud representations, enabling efficient handling of netlist with up to hundreds of thousands to millions nodes. All types of data, including netlist files and image data, are encoded into latent space as features and fed into the model for static voltage drop prediction. This enables the integration of data from multiple modalities for complementary predictions. Experimental results demonstrate that our proposed algorithm can achieve the best F1 score and the lowest MAE among the winning teams of the ICCAD 2023 contest and the state-of-the-art algorithms.
arXiv:2602.09388v2 Announce Type: replace
Abstract: Multilingual pre-trained language models(mPLMs) offer significant benefits for many low-resource languages. To further expand the range of languages these models can support, many works focus on continued pre-training of these models. However, few works address how to extend mPLMs to low-resource languages that were previously unsupported. To tackle this issue, we expand the model's vocabulary using a target language corpus. We then screen out a subset from the model's original vocabulary, which is biased towards representing the source language(e.g. English), and utilize bilingual dictionaries to initialize the representations of the expanded vocabulary. Subsequently, we continue to pre-train the mPLMs using the target language corpus, based on the representations of these expanded vocabulary. Experimental results show that our proposed method outperforms the baseline, which uses randomly initialized expanded vocabulary for continued pre-training, in POS tagging and NER tasks, achieving improvements by 0.54% and 2.60%, respectively. Furthermore, our method demonstrates high robustness in selecting the training corpora, and the models' performance on the source language does not degrade after continued pre-training.
arXiv:2602.09464v2 Announce Type: replace
Abstract: Vericoding refers to the generation of formally verified code from rigorous specifications. Recent AI models show promise in vericoding, but a unified methodology for cross-paradigm evaluation is lacking. Existing benchmarks test only individual languages/tools (e.g., Dafny, Verus, and Lean) and each covers very different tasks, so the performance numbers are not directly comparable. We address this gap with AlgoVeri, a benchmark that evaluates vericoding of $77$ classical algorithms in Dafny, Verus, and Lean. By enforcing identical functional contracts, AlgoVeri reveals critical capability gaps in verification systems. While frontier models achieve tractable success in Dafny ($40.3$% for Gemini-3 Flash), where high-level abstractions and SMT automation simplify the workflow, performance collapses under the systems-level memory constraints of Verus ($24.7$%) and the explicit proof construction required by Lean (7.8%). Beyond aggregate metrics, we uncover a sharp divergence in test-time compute dynamics: Gemini-3 effectively utilizes iterative repair to boost performance (e.g., tripling pass rates in Dafny), whereas GPT-OSS saturates early. Finally, our error analysis shows that language design affects the refinement trajectory: while Dafny allows models to focus on logical correctness, Verus and Lean trap models in persistent syntactic and semantic barriers. All data and evaluation code can be found at https://github.com/haoyuzhao123/algoveri.
arXiv:2602.09795v2 Announce Type: replace
Abstract: Schelling's model of segregation demonstrates that even in the absence of social or governmental interventions, individuals with mild in-group preferences can self-organize into strongly segregated neighborhoods. Many variants of this celebrated model have been proposed by assuming agents tend to increase their satisfaction. Complementary to this traditional, utility-based approach, we model residential moves using satisfaction-independent reaction rates in a spatially extended chemical reaction network. The resulting model exhibits an emergent phenomenon: despite symmetric in-group preferences, the system undergoes a tipping transition at a critical preference level, beyond which one agent type dominates. We characterize this asymmetric phase transition in details using mean-field analysis, numerical simulations and finite size scaling methods. We find that while the transition shares key features with the Ising universality class, such as $\mathbb{Z}_2$ symmetry breaking and similar exponent ratios, the full set of critical exponents does not match any known universality class.
arXiv:2602.12147v4 Announce Type: replace
Abstract: Time series foundation models (TSFMs) are revolutionizing the forecasting landscape from specific dataset modeling to generalizable task evaluation. However, we contend that existing benchmarks exhibit common limitations in four dimensions: constrained data composition dominated by reused legacy sources, compromised data integrity lacking rigorous quality assurance, misaligned task formulations detached from real-world contexts, and rigid analysis perspectives that obscure generalizable insights. To bridge these gaps, we introduce TIME, a next-generation task-centric benchmark comprising 50 fresh datasets and 98 forecasting tasks, tailored for strict zero-shot TSFM evaluation free from data leakage. Integrating large language models and human expertise, we establish a human-in-the-loop benchmark construction pipeline to ensure high data integrity and redefine task formulation by aligning forecasting configurations with real-world operational requirements and variate predictability. Furthermore, we propose a novel pattern-level evaluation perspective that moves beyond traditional dataset-level evaluations based on static meta labels. By leveraging structural time series features to characterize intrinsic temporal properties, this approach offers generalizable insights into model capabilities across diverse patterns. We evaluate 12 TSFMs and establish a multi-granular leaderboard to facilitate in-depth analysis and visualized inspection. The leaderboard is available at https://huggingface.co/spaces/Real-TSF/TIME-leaderboard.
arXiv:2602.12643v2 Announce Type: replace
Abstract: We present Unified Latent Dynamics (ULD), a novel reinforcement learning algorithm that unifies the efficiency of model-free methods with the representational strengths of model-based approaches, without incurring planning overhead. By embedding state-action pairs into a latent space in which the true value function is approximately linear, our method supports a single set of hyperparameters across diverse domains -- from continuous control with low-dimensional and pixel inputs to high-dimensional Atari games. We prove that, under mild conditions, the fixed point of our embedding-based temporal-difference updates coincides with that of a corresponding linear model-based value expansion, and we derive explicit error bounds relating embedding fidelity to value approximation quality. In practice, ULD employs synchronized updates of encoder, value, and policy networks, auxiliary losses for short-horizon predictive dynamics, and reward-scale normalization to ensure stable learning under sparse rewards. Evaluated on 80 environments spanning Gym locomotion, DeepMind Control (proprioceptive and visual), and Atari, our approach matches or exceeds the performance of specialized model-free and general model-based baselines -- achieving cross-domain competence with minimal tuning and a fraction of the parameter footprint. These results indicate that value-aligned latent representations alone can deliver the adaptability and sample efficiency traditionally attributed to full model-based planning.
arXiv:2602.16966v2 Announce Type: replace
Abstract: Scalable methods for networked multi-agent reinforcement learning let each agent plan using only a small neighborhood of the agent graph. This works only when the system is value-local, meaning a perturbation at one agent affects the long-run value at another agent weakly when the two are far apart. In the average-reward setting, the standard way to certify locality is the Dobrushin row-sum bound on a single matrix $C^\pi$ that captures how each agent's next state depends on each other agent's current state. To make this matrix easy to work with, prior work bounds it by a supremum over joint actions. The resulting bound is independent of the policy, but it is loose whenever the policy never picks the worst-case action. We split $C^\pi$ into pieces that separately track environment sensitivity and policy sensitivity, $C^\pi \preceq E^{\mathrm s}+E^{\mathrm a}\Pi(\pi)$, where $E^{\mathrm s}$ measures how the next state moves with the current state, $E^{\mathrm a}$ measures how it moves with the current action, and $\Pi(\pi)$ measures how reactive the policy is to changes in state. The spectral radius of $H^\pi := E^{\mathrm s}+E^{\mathrm a}\Pi(\pi)$ then controls the decay of the average-reward Poisson solution, and the spectral certificate $\rho(H^\pi)<1$ is strictly weaker than the row-sum condition $\|H^\pi\|_\infty<1$ on the same matrix and applies in regimes where policy-independent action-supremum bounds used in prior Dobrushin-style work cannot. For temperature-$\tau$ softmax policies we get $\Pi(\pi)\le L/(2\tau)$, so the softmax temperature directly controls locality. We use this decay result to give a deterministic oracle guarantee for a block-coordinate KL-proximal policy-improvement template whose truncation bias decays exponentially in the message-passing radius $\kappa$.
arXiv:2602.17434v2 Announce Type: replace
Abstract: Multi-agent planning under Signal Temporal Logic (STL) is often hindered by collaborative tasks that lead to computational challenges due to the inherent high dimensionality of the problem, preventing scalable synthesis with satisfaction guarantees. To address this, we formulate STL planning as an optimization program under multi-agent STL constraints and introduce a penalty-based unconstrained relaxation that can be efficiently solved via a Block-Coordinate Gradient Descent (BCGD) method, where each block corresponds to a single agent's decision variables, thereby mitigating complexity. By utilizing a quadratic penalty function defined via smooth STL semantics, we show that BCGD iterations converge to a stationary point of the penalized problem under standard regularity assumptions. To enforce feasibility, the BCGD solver is embedded within a two-layer optimization scheme: inner BCGD updates are performed for a fixed penalty parameter, which is then increased in an outer loop to progressively improve multi-agent STL robustness. The proposed framework enables scalable computations and is validated through various complex multi-robot planning scenarios.
arXiv:2602.20651v3 Announce Type: replace
Abstract: In modern applications such as ECG monitoring, neuroimaging, wearable sensing, and industrial equipment diagnostics, complex and continuously structured data are ubiquitous, presenting both challenges and opportunities for functional data analysis. However, existing methods face a critical trade-off: conventional functional models are limited by linearity, whereas deep learning approaches lack interpretable region selection for sparse effects. To bridge these gaps, we propose a sparse Bayesian functional deep neural network (sBayFDNN). It learns adaptive functional embeddings through a deep Bayesian architecture to capture complex nonlinear relationships, while a structured prior enables interpretable, region-wise selection of influential domains with quantified uncertainty. Theoretically, we establish rigorous approximation error bounds, posterior consistency, and region selection consistency. These results provide the first theoretical guarantees for a Bayesian deep functional model, ensuring its reliability and statistical rigor. Empirically, comprehensive simulations and real-world studies confirm the effectiveness and superiority of sBayFDNN. Crucially, sBayFDNN excels in recognizing intricate dependencies for accurate predictions and more precisely identifies functionally meaningful regions, capabilities fundamentally beyond existing approaches.
arXiv:2502.05349v2 Announce Type: replace-cross
Abstract: Two-stage stochastic programs (2SPs) are widely used for decision-making under uncertainty, but their practical deployment is often limited by the large number of scenarios needed to approximate the conditional distribution of uncertain outcomes. We study contextual scenario generation: given contextual information, learn to produce a small, user-specified set of surrogate scenarios that, when used as input into the 2SP, lead to high-quality 2SP decisions. Existing scenario generation methods either ignore contextual information or are computationally burdensome in this setting. We propose contextual scenario generation (CSG), which learns a mapping from context to a set of surrogate scenarios. We develop two complementary methodologies: (i) a distributional approach that learns a mapping from context to scenarios by minimizing a kernel-based distance to the conditional distribution, and (ii) a task-based approach that selects the mapping to optimize decision quality via differentiating through a learned surrogate of the downstream 2SP objective. Both approaches are broadly applicable and require only repeated solution of the underlying subproblems and 2SPs defined on the generated scenarios. We provide finite-sample generalization guarantees and demonstrate strong empirical performance across multiple 2SP classes.
arXiv:2606.04127v1 Announce Type: new
Abstract: Medical question answering is a high-stakes setting where factual errors can have serious consequences. Retrieval-augmented generation (RAG) is widely viewed as a promising solution, and prior work has reported substantial gains for large medical QA models. We revisit this assumption across a broad range of open-weight instruction-tuned models spanning 7B to 72B parameters. Across five models, ten biomedical QA datasets, four retrieval methods, and four retrieval corpora, we find that retrieval yields only small and inconsistent improvements over a no-retrieval baseline, typically within 1-2 points. In contrast, the choice of backbone model has a much larger effect than the choice of retriever or corpus, and expert and layman retrieval sources perform similarly in most settings. These results suggest that the main bottleneck is not retrieval quality alone, but the model's limited ability to use retrieved evidence effectively.
arXiv:2411.05591v2 Announce Type: replace-cross
Abstract: We systematically study several network-based Expectation-Maximization (EM) algorithms for the Gaussian mixture model within decentralized federated learning (DFL). Our theoretical investigation shows that directly extending the classic EM algorithm to DFL leads to a biased estimator when data are heterogeneously distributed across sites. To address this, we introduce a momentum network EM (MNEM) algorithm, which integrates information from both current and historical estimators from previous DFL iterations. We further develop a semi-supervised MNEM (semi-MNEM) algorithm, which utilizes information provided by partially labeled data. Rigorous theoretical analysis demonstrates that the MNEM estimator can achieve the same asymptotic efficiency as the whole-sample estimator under appropriate regularity conditions, even with heterogeneous data. Moreover, the semi-MNEM estimator significantly improves the convergence speed of the MNEM algorithm, even if different mixture components are poorly separated. Extensive simulations are conducted, and a widely used chest X-ray dataset is analyzed to demonstrate the finite-sample performance of the proposed methods.
arXiv:2408.02548v3 Announce Type: replace-cross
Abstract: We determine all the Betti numbers of the $q$-ary second order Reed-Muller codes of length $q^2$, and also of the elongations of matroids associated to these codes. We then use it to determine the higher weight spectra of these codes. As a special case, we recover some results of Kaplan and Matei about counting certain curves over finite fields with prescribed rational intersection points. In geometric terms, our results relate to the affine Veronesean by which we mean the image of the affine plane $\mathbb{A}^2$ under the quadratic Veronese embedding of $\mathbb{P}^2$ in $\mathbb{P}^5$. Indeed, finding the higher weight spectra, of the Reed-Muller code considered here, corresponds to determining the number of $\mathbb{F}_q$-rational points on all possible sections of this affine Veronesean by linear subvarieties of $\mathbb{P}^5$.
arXiv:2205.08609v3 Announce Type: replace-cross
Abstract: Climate and environmental applications increasingly rely on high-dimensional prediction from remote sensing and other scientific data. Neural networks (NN) can deliver strong accuracy in these settings, but they are often hard to audit and hard to align with domain knowledge. As an alternative, we propose bagged polynomial regression with random projections (BPR), an econometrics-native ensemble that averages many regularized low-degree polynomial models fit on randomly selected covariate groups. We provide novel finite-sample and asymptotic risk bounds and show how covariate partitioning can improve rates for smooth target functions by controlling dictionary basis growth. Rate improvements may be particularly relevant for the estimation of marginal effects. In an application to satellite-based crop classification using optical and radar imagery, BPR matches NN accuracy while remaining straightforward to diagnose. We provide practical transparency tools, coefficient summaries and partial-dependence diagnostics, that show BPR captures intuitive feature relationships that NNs do not.
arXiv:2602.20971v3 Announce Type: replace
Abstract: Bubeck and Selke (2021) propose the connection between the Law of Robustness and robust generalization error as an open problem. The Law of Robustness states that overparameterization is necessary for models to interpolate robustly, i.e., the interpolating function is required to be Lipschitz. Wu et al. (2023) extend this law to arbitrary data distributions, proving that the Lipschitz constant satisfies $L = \Omega(n^{1/d})$. Robust generalization, on the other hand, asks whether small robust training loss implies small robust test loss. This can be studied using statistical learning techniques such as Rademacher complexities, where a bound on the Rademacher complexity of the robust loss class implies a bound on the Lipschitzness of the function class. We use this connection to explicitly link the two for arbitrary data distributions. (i) We prove that the order of the Lipschitz bound remains the same when considering the global Rademacher complexity of robust loss classes. (ii) At the local scale, i.e., for subsets of functions with small empirical error, the order of the Lipschitz bound changes with the perturbation radius $\rho$ and the localized concentration term $\sqrt{r/n}$.
arXiv:2606.03606v2 Announce Type: replace
Abstract: Large language models achieve strong performance on arithmetic reasoning benchmarks, and one common response to arithmetic brittleness is to delegate computation to code. Yet models are still often used in settings where they must reason directly from natural language, and trustworthy models should solve small-number arithmetic word problems without external tools. Prior work shows that LLMs are sensitive to numerical variation: a model may solve an original problem but fail on structurally similar variants requiring the same reasoning procedure with different numbers. We ask whether this fragility persists under a stricter setting involving small, schema-preserving numeric changes that retain the original reasoning program and avoid large-number stress tests. We introduce an automatic algorithm for generating numeric-remapping attacks on arithmetic word problems. Unlike template-based perturbation methods requiring manual schemas or constraints, our approach derives problem-specific symbolic representations, generates constrained numeric remappings, recomputes gold answers, and realizes transformed questions through deterministic edits guided by LLM-generated edit plans. Stage-wise validation and a high-confidence audit retain reliable attacks, making the pipeline scalable with limited human intervention. We evaluate DeepSeek-R1 (70B), Gemma4 (31B), and GPT-OSS (120B) on GSM8K, MAWPS, and MultiArith. On GSM8K, completed runs show conditional accuracy drops of 12.16 to 25.82 percentage points. MAWPS and MultiArith are far more stable, with most attacked accuracies near or above 98%. These results show that numeric-remapping robustness depends strongly on dataset structure: GSM8K remains sensitive even when reasoning programs are preserved and answers are recomputed, while shorter, more regular datasets are more robust.
arXiv:2606.03307v2 Announce Type: replace
Abstract: Graph foundation models (GFMs) emerged as a dominant paradigm in graph representation learning by leveraging large-scale pre-training for cross-domain inference. However, the parameterized knowledge encoded within these models is insufficient to cope with distribution shifts, limiting their generalization ability. To mitigate this issue, retrieval-augmented generation (RAG) has been introduced to incorporate external knowledge at inference time. Nevertheless, existing RAG frameworks operating in Euclidean space suffer from a fundamental geometric limitation: the polynomial volume growth of Euclidean space is inherently mismatched with the tree-structured external knowledge bases. This mismatch leads to the loss of semantic granularity in retrieval and gives rise to the hubness phenomenon.To address this limitation, we propose a Hyperbolic Retrieval-Augmented Generation (HyRAG) framework designed to enhance the generalization capabilities of GFMs. Specifically, the introduced Hyperbolic Knowledge Indexing module retains the tree-like hierarchies of the external knowledge base by modeling them within hyperbolic space. The Multi-granularity Retrieval module then provides GFMs with the global semantic anchors and local semantic nuances through coarse-grained and fine-grained knowledge retrieval, respectively. Finally, the Dual-path Fusion module achieves effective knowledge integration for graph tasks at both the feature and structural levels. Experiments on multiple graph benchmarks demonstrate significant improvements in the zero-shot setting, highlighting the generalization of our method for robust GFMs inference.
arXiv:2606.03201v2 Announce Type: replace
Abstract: Reinforcement learning from expert videos across visually distinct domains is challenging due to the absence of reward signals and the presence of domain gaps. We introduce XIPER (Cross-domain Video Prediction Reward), a reward model for learning from expert videos collected in a visually different domain, where the agent's appearance differs due to factors such as color, morphology, or the sim-to-real gap. More specifically, XIPER trains a cross-domain video prediction model that maps agent observations into the expert domain and uses the prediction likelihood as a reward signal. Experiments on the DMC Color Suite (8 tasks) and DMC Body Suite (3 tasks) show that XIPER consistently outperforms baselines despite domain gaps such as differences in agent color and morphology. We further analyze XIPER on a sim-to-real transfer dataset, demonstrating that it produces meaningful reward signals for real-robot observations given only simulated expert videos. Code, pretrained models, datasets and video demonstrations can be found on our project webpage: https://sites.google.com/view/xiper
arXiv:2606.03175v2 Announce Type: replace
Abstract: Instance Goal Navigation (IGN) requires an embodied agent to find a specific object instance among distractors from an under-specified natural-language description. Such ambiguity often cannot be resolved from perception and language alone, making interaction with an oracle a natural mechanism for disambiguation. Prior interactive methods allow oracle queries but treat lightweight clarification and route-level guidance alike, letting agents boost success rate through repeated high-information questions rather than by resolving the underlying ambiguity efficiently. We recast interactive IGN as a cost-sensitive uncertainty-reduction problem, where the agent should ask the question whose answer provides the largest reduction in navigation uncertainty relative to its penalty. To this end, we apply an information-gain analysis on existing navigation corpora to identify which cues reduce navigation uncertainty, yielding a compact set of question types and data-derived weights. However, existing interactive navigation benchmarks do not model the cost of different question types or evaluate how efficiently agents use interaction, making them unsuitable for studying cost-sensitive interaction. Based on this taxonomy, we construct a benchmark for diagnosing interaction behavior and efficiency, together with a Weighted Success Rate metric that penalizes each query by its derived cost. We further propose a zero-shot MLLM navigator that selectively queries at each decision step only when the expected uncertainty reduction justifies the interaction cost.
arXiv:2601.07036v2 Announce Type: replace
Abstract: Hybrid reasoning language models are commonly controlled through high-level Think/No-think instructions to regulate reasoning behavior, yet we found that such mode switching is largely driven by a small set of trigger tokens rather than the instructions themselves. Through attention analysis and controlled prompting experiments, we show that a leading ``Okay'' token induces reasoning behavior, while the newline pattern following ``</think>'' suppresses it. Based on this observation, we propose Mid-Think, a simple training-free prompting format that combines these triggers to achieve intermediate-budget reasoning, consistently outperforming fixed-token and prompt-based baselines in terms of the accuracy-length trade-off. Furthermore, applying Mid-Think to RL training after SFT reduces training time by approximately 15% while improving final performance of Qwen3-8B on AIME from 69.8% to 72.4% and on GPQA from 58.5% to 61.1%, demonstrating its effectiveness for both inference-time control and RL-based reasoning training.
arXiv:2602.07699v2 Announce Type: replace
Abstract: We describe a series of experiments done using a commercially available optical pumping apparatus that is currently being used in physics teaching labs at over one hundred universities. Our focus here is to provide an extensive and detailed examination of the capabilities of this instrument, including numerous examples of measurements and data analysis, presented as a supplement to the manufacturers user manual. Our hope is that instructors using this or similar optical pumping instruments will find the experiments described here useful for designing and implementing the curricula in their own physics teaching labs.