We study multigrade deep learning (MGDL) as a principled framework for structured error refinement in deep neural networks. While the approximation power of neural networks is now relatively well understood, training very deep architectures remains challenging due to highly nonconvex and often ill-conditioned optimization landscapes. In contrast, for relatively shallow networks, most notably certain one-hidden-layer ReLU models, training admits convex reformulations with global guarantees under appropriate settings, motivating learning paradigms that improve stability while scaling to depth. MGDL builds on this insight by training deep networks grade by grade: previously learned grades are frozen, and each newly added grade-wise subnetwork is composed on top of the previously learned grades and trained to fit the residual left by the current approximation, yielding a structured and interpretable hierarchical refinement process. We develop an operator-theoretic foundation for MGDL and prove that, for any continuous target function defined on a hypercube, there exists a fixed-width multigrade ReLU scheme whose residuals are pointwise nonincreasing in magnitude and converge uniformly to zero, with strict $L^p$-norm decay at every nontrivial grade for $p\in [1,\infty)$. To the best of our knowledge, this work provides the first rigorous constructive approximation guarantee showing that a grade-wise residual refinement scheme can achieve vanishing error in a fixed-width multigrade ReLU architecture.
Muhammad Rajabinasab, Michael E. Houle, Oussama Chelly, Arthur Zimek
Many novel unsupervised feature selection methods are proposed each year, yet their empirical evaluation is limited to supervised and unsupervised evaluation metrics computed on selected datasets, along with comparisons to existing methods. However, in the absence of an established evaluation baseline, it is difficult to determine the value added to the existing literature by each of these methods, and how effective their underlying approaches are. In this evaluation study, we are using random feature selection as a baseline for evaluating unsupervised feature selection methods. While a random selection of features might seem to be a rather weak baseline, we show in this evaluation study that many of the state-of-the-art methods in unsupervised feature selection are outperformed by random feature selection in both performance and efficiency. Accordingly, we emphasize the strict requirement of considering random feature selection as a baseline in the development process of novel unsupervised feature selection methods, and we encourage the community to develop better evaluation measures for feature selection. Both would help to ensure consistent improvement and realistic assessment.
Dan Assouline, Erwan Koch, Federico Amato, Filippo Quarenghi, Daniele Nerini, Thibaut Loiseau, Kyle van de Langemheen, Tom Beucler
Skillful medium-range precipitation forecasting at kilometer scale remains challenging over complex terrain because precipitation arises from multiscale nonlinear processes that global models cannot explicitly resolve at affordable cost. Global AI weather models can produce skillful medium-range forecasts, but their native 0.25 degrees resolution limits direct use for local hazard applications. Statistical downscaling can help bridge this gap, yet existing approaches often struggle with state-dependent, and especially lead-time-dependent, biases in global forecasts. We introduce SwAIther-Precip, a lead-time-aware downscaling framework that converts coarse-resolution AIFS forecasts into probabilistic km-scale precipitation fields over Switzerland. First, a U-Net conditioned on lead time via feature-wise linear modulation deterministically corrects systematic biases at coarse resolution. This targeted correction enables a cheaper super-resolution stage conditioned only on corrected precipitation, allowing direct training on observations rather than on the full atmospheric state. A diffusion-based model then generates fine-scale spatial variability independently of lead time. Using AIFS forecasts and CombiPrecip radar-gauge observations, SwAIther-Precip reduces CRPS by 48% relative to raw AIFS. The generated fields reproduce observed spatial variability with spectral fidelity above 0.85 at large scales and 0.88 at small scales, corresponding to an effective resolution of approximately 4 km on a 1 km grid for lead times up to 5 days. Training across lead times further improves long-range performance, yielding a 13% CRPS reduction at 6 days relative to lead-time-specific models. These results show that explicitly correcting lead-time-dependent biases before generative super-resolution is key to efficient km-scale probabilistic downscaling of global AI precipitation forecasts.
Jeffrey Qin, Zhuangfei Gao, Wasu Top Piriyakulkij, Mia Radovanovic, Jessica Sommerville, Kevin Ellis, Marta Kryven
Children expertly construct mental models under uncertainty despite limited world knowledge. Which computational principles make this possible? We study this question in an interactive physical Box Task in which participants infer a latent cause through interaction with a stochastic and partially observed environment. The Box Task was originally designed to study children's puzzle-solving qualitatively. We contribute a computational analysis of behavior that formalizes the cognitive process by which children generate new hypotheses and respond to environmental uncertainty as inference unfolds. Our modeling framework is based on Bayesian particle-based inference, which maintains a small set of working hypotheses and continually generates new ones. We give two complementary implementations of hypothesis representations within this framework: (1) as constraints on evidence, and (2) as executable programs evaluated against evidence. Using the constraint-based formulation, we show that the model needs to account both for subjective reliability of evidence and online hypothesis generation to explain key signatures of children's behavior. Using the programmatic formulation, we show that hypotheses may take the form of partial compositional rules to guide information gathering when evidence interpretation is hard. The programmatic hypotheses are implemented using LLM-based program synthesis, evaluated across three prompt variants and several LLM backends, demonstrating robustness of our approach. Together, these results extend current models of children's inference to account for subjective reliability of evidence and open-ended hypothesis spaces, and provide a computational analysis of the form generated hypotheses may take in practice.
Earth Observation (EO)-based agricultural monitoring is shifting from pixel-level spectral analysis toward representation-driven modeling, yet systematic evaluation of learned representations remains limited. Geospatial foundation models (FMs) generate high-dimensional embeddings encoding complex spatio-temporal land-surface properties, but their utility and interpretability for agricultural applications remain insufficiently understood. This study presents a representation-agnostic framework evaluating a 64-dimensional AlphaEarth embedding space against Sentinel-2 reflectance, NDVI phenology, and harmonic features across geometric separability, temporal dynamics, spatial robustness, and cross-regional generalisation using field-scale data from the Conterminous United States (CONUS; 2017-2023) and Northern India (2021). Traditional features achieve strong linear separability (Fisher ratio), while embeddings provide superior distribution-aware separation - Bhattacharyya and Mahalanobis distances up to an order of magnitude higher for spectrally similar pairs (e.g. Corn/Soybeans: 3.9 vs. 0.10). Embeddings uniquely capture multi-year crop rotation dynamics, yielding a separability ratio of 1.784, versus a maximum of 1.003 for all engineered features. Under spatially explicit validation, embeddings exhibit lower residual spatial autocorrelation of errors (mean Moran's $I \approx 0.17$), reducing reliance on geographic proximity. However, harmonic features show stronger cross-continental transferability (US $\to$ India Macro F1: 0.493 vs. 0.326 for embeddings), with NDVI showing a similar advantage. The combined feature space yields balanced performance. These findings reveal a trade-off between representational richness and cross-region transferability, providing a framework for feature selection in large-scale crop classification.
Dongyoon Hahm, Woogyeol Jin, Dennis Wei, Yi Zhou, Nathalie Baracaldo, Kimin Lee
Fine-tuning large language models (LLMs) can induce behaviors beyond those explicitly targeted. For example, training a model to respond more warmly can make it more sycophantic, and training it to write insecure code can cause broad misalignment on prompts unrelated to coding. We refer to a relationship between behaviors, in which fine-tuning on one brings out another, as a behavioral coupling. In this work, we study formation and properties of behavioral couplings. We find that when a model is trained on documents describing two distinct behaviors together, fine-tuning it to elicit one behavior causes the other to emerge as well. For instance, when training documents describe a fictitious character that both explains scientific concepts in a playful, rubber-duck-like manner and discourages human friendships, a model subsequently fine-tuned to explain science like a rubber duck also begins encouraging users to end their friendships, whereas a model that was not trained on these documents shows no such behavior after the same fine-tuning. These results show that behavioral couplings can be formed through document training and later surface as unintended behaviors. More concerningly, we find that couplings can compose. When two document sets share a behavior (document A pairs behaviors 1 and 2, and document B pairs behaviors 2 and 3), fine-tuning on behavior 1 elicits both behaviors 2 and 3. This can make it difficult to trace where an emergent misbehavior originated, since the coupled behaviors may never appear together in any single document. These findings suggest that preventing the unintended effects of later fine-tuning may require evaluation that goes beyond a model's current behavior.
Heejun Kim, Junyoung Lee, SangLyul Cho, Dongsu Han, Insu Han, Sehoon Kim
Looped Transformers improve parameter efficiency by applying shared Transformer blocks over recurrent loops, increasing computational depth without increasing the parameter count. However, KV cache memory still scales with loop count, becoming a key memory bottleneck that limits batch size and inference throughput. KV cache quantization can alleviate this bottleneck, but existing methods often suffer substantial accuracy degradation at aggressive low-precision regimes. We observe that looped Transformers offer a unique opportunity: KV states across loops are highly similar. Based on this observation, we propose ResidualQuant, which uses the final-loop KV states as a reference and represents the remaining loops with low-precision residuals. Our method further combines least-square scaling and rotation-based quantization, as well as loop-wise mixed precision, to enable accurate quantization down to INT2 while retaining efficient reconstruction. Across multiple looped Transformer models and mathematical reasoning and code generation benchmarks, ResidualQuant consistently improves the accuracy-memory tradeoff over state-of-the-art rotation-based KV quantization. In particular, our method retains accuracy close to BF16 under mixed-precision settings while reducing theoretical KV storage by 80.7%, achieving up to 13.0% higher accuracy than the rotation-based baseline at nearly the same memory budget. On an RTX 5090, reduced KV memory traffic improves fixed-batch decode throughput by up to 3.66x, while the smaller memory footprint enables up to 8x larger batches, improving peak throughput by up to 5.38x.
Flow Matching has recently gained attention in generative modeling as a simple and flexible alternative to diffusion models. While existing statistical guarantees adapt tools from the analysis of diffusion models, we take a different perspective by connecting Flow Matching to kernel density estimation. We first verify that the kernel density estimator matches the optimal rate of convergence in Wasserstein distance up to logarithmic factors, improving existing bounds for the Gaussian kernel. Based on this result, we prove that for sufficiently large networks, Flow Matching achieves the optimal rate up to logarithmic factors. If the target distribution lies on a lower-dimensional manifold, we show that the kernel density estimator profits from the smaller intrinsic dimension on a small tube around the manifold. The faster rate also applies to Flow Matching, providing a theoretical foundation for its empirical success in high-dimensional settings.
Closed-loop AI scientists can generate candidate designs at low marginal computational cost, whereas reliable feedback may require wet-lab synthesis, characterization, or high-fidelity computation. Addressing this imbalance through custom laboratory automation remains infrastructure-intensive and costly, while replacing new experiments with a fixed surrogate leaves persistent model errors that can be amplified by optimization. We propose \emph{online surrogate repair} (OSR), a closed-loop algorithm that uses sparse high-fidelity evaluations to update the surrogate throughout a longer agent search conducted primarily with inexpensive surrogate feedback. An acquisition rule selects which designs from the agent's accumulated proposals receive high-fidelity evaluation, and the resulting labels update the surrogate used in subsequent episodes. Across controlled synthetic environments, we demonstrate that improving global surrogate fit does not necessarily reduce maximum regret, whereas Q90-UCB and expected improvement (EI) substantially reduce regret by directing evaluations toward regions that determine the optimizer's decisions. \osradded{Across LLM and Chemeleon workflows, Online EI reduces the amount of oracle queries required by 84.3--90.3\% for matching discovery counts and by 77.9--81.6\% for matching the quality of the best material found.} Online surrogate repair introduces a third feedback regime between fixed-surrogate operation and high-fidelity feedback after every episode, separating the frequency of high-fidelity evaluation from the duration of the agent's search.
This paper proposes a method for vertex-wise aggregation sampling of a broad class of graph signals, designed to attain the best possible recovery based on the generalized sampling theory. This is achieved by designing a sampling operator by an optimization problem, which is inherently non-convex, as the best possible recovery imposes a rank constraint. An existing method for vertex-wise aggregation sampling is able to control the number of active vertices but cannot incorporate prior knowledge of mandatory or avoided vertices. To address these challenges, we formulate the operator design as a problem that handles a constraint on the number of active vertices and prior knowledge on specific vertices for sampling, mandatory inclusion or exclusion. We transformed this constrained problem into a difference-of-convex (DC) optimization problem by using the nuclear norm and a DC penalty for vertex selection. To solve this, we develop a convergent solver based on the general double-proximal gradient DC algorithm. The effectiveness of our method is demonstrated through experiments on various graph signal models, including real-world data, showing superior performance in the recovery accuracy compared to existing methods.
Despite their practical significance, modern AI systems pose significant societal risks, including toxicity, malicious use, and misinformation. Their mitigation depends not only on technical feasibility but also on the incentives of firms that develop and deploy these systems. We show how profit maximization shapes safety investment in an AI supply chain with an upstream LLM provider and several competing downstream firms. In our game-theoretic model, we derive the market-driven level of safety investment and compare it with the levels that maximize industry profit and market welfare. We find that downstream competition generally induces downstream firms to invest more in safety than the industry-profit-optimal level. By contrast, the upstream provider may underinvest relative to these benchmarks because it is not fully compensated for the surplus from safety investments. Finally, we analyze how different regulatory interventions affect the participants' incentives and profits, and the resulting levels of model safety.
Zijuan Zhao, Zheren Fu, Hou Xia, Licheng Zhang, Yi Liu, Zhendong Mao
Assessing whether multimodal content aligns with macro-societal values, such as peace, justice, and freedom, has become an increasingly urgent challenge. Existing frameworks are largely confined to safety-oriented taxonomies, text-only psychometric probes, or single-label classification. Therefore, we propose MAVEN, a hierarchical framework for macro-societal value evaluation of multimodal content, grounded in international human-rights instruments and cultural value theory. MAVEN organizes values into 6 primary dimensions and 72 secondary indicators, supporting multi-level quantitative scoring. Building on MAVEN, we construct a human-verified multimodal benchmark and a soft-match metric to evaluate VLMs' assessments across value dimensions. For evaluator optimization, we propose a span-adaptive variant of multi-level preference optimization for evaluator distillation, together with a training-free multi-role consensus strategy at inference time. We evaluate existing open- and closed-source VLMs on our benchmark, revealing shared tendencies and clear differences in macro-societal value judgments. Experiments show that our compact 2B evaluator matches its 8B counterpart in the same family and approaches frontier closed-source VLMs, offering a practical path toward scalable macro-societal value evaluation. Our SA-MDPO implementation and MacroValue-Bench are available at https://github.com/zzzzzzzzjj/MAVEN.
Elias Sandmann, Sebastian Lapuschkin, Wojciech Samek
Recent mechanistic work has uncovered learned algorithms within neural networks, from modular arithmetic to search and planning in game-playing agents. But does algorithmic structure guarantee algorithmic behavior? We investigate this in Leela Chess Zero, the strongest neural chess engine, where prior work identified learned look-ahead. By extending the logit lens to its move-selecting policy network, we discover that correct puzzle solutions--including immediate checkmates--often appear in intermediate layers but are systematically overridden in the final output, a phenomenon we term "forgotten puzzles". Replicating prior analyses on these positions, we find that look-ahead operates normally--future moves of the correct continuation are represented, causally important, and linearly decodable--ruling out a failure of the algorithm. Instead, late layers increasingly shift toward safe play over aggression. To test whether this shift drives the override, we steer the model against these preferences and recover 71.7% of forgotten puzzles, providing causal evidence that safety priors override algorithmically computed solutions. These findings show that algorithmic structure does not guarantee algorithmic behavior: a model can internally solve a problem and still output the wrong answer.
Accurate workload forecasting is critical for elastic resource provisioning in web-scale cloud services, where distribution shifts driven by viral content, product launches, and user behavior degrade offline-trained models rapidly. Naive online learning recovers accuracy but incurs prohibitive per-step compute cost. We propose AdaptLSTM, an adaptive online framework that detects drift via validation-calibrated thresholds and applies selective, targeted updates. On the Alibaba Machine Trace, AdaptLSTM recovers 54% of Naive Online's improvement at 20% cost ($2.7\times$ efficiency, $p=0.002$ over 10 seeds). On the more volatile Container Trace, it achieves 96% at 20% cost ($4.8\times$ efficiency, $+75\%$ MAE reduction over Static). Unlike classical drift detectors (ADWIN, DDM, Page-Hinkley) which fail to trigger on regression-scale error streams, AdaptLSTM fires 42 times over 301 steps and outperforms matched-budget baselines. Wall-clock profiling shows $1.33\times$ throughput gain and 45% update-time reduction. The framework is model-agnostic: identical Pareto patterns hold for LSTM, GRU, and Transformer backbones.
In reinforcement learning with verifiable rewards (RLVR), whether entropy rises or falls can depend on choices that matter little under exact sampling, such as whether over-length responses are masked or counted as failures. We trace this to sampler mismatch: rollouts are drawn with top-k/top-p truncation or at a temperature the learner does not apply. We prove that a constant advantage shift then changes the entropy of a softmax policy, at first order and in expectation, with a sign fixed by the sampler. Positive shifts sharpen the policy under any sampler whose likelihood ratio to the policy increases with token probability, including top-k, top-p, min-p, and temperatures below one, and flatten it under flatter samplers. In group-based RLVR with token-level losses, the baseline-dependent part of this drift is set by the token-weighted net advantage, which masking and failing can push to opposite signs. The rest comes from a covariance between advantages and the sampler's sharpness gap and feeds back on itself: in a two-level model, a push away from a sharper sampler takes off in finite time, while a pull toward it saturates. On a 2B model, the entropy slope is linear in shifts of the net advantage under truncated sampling and indistinguishable from flat under exact sampling. Token-level importance ratios leave 98% of the drift, and four clipped minibatch updates per batch amplify it sevenfold. In 100-step runs, a net-zero baseline only delays an entropy explosion, and entropy raised at the start keeps rising only under truncation; cancelling the first-order drift at every step keeps entropy bounded without changing the learner. On a 9B reasoning model forked from an uncorrected run, zeroing the net advantage or renormalizing the learner raises held-out accuracy after ten steps by 8.2 and 8.8 points over the uncorrected rule, with no difference under exact sampling.
Jaehun Shon, Jinha Choi, Jongwook Jeon, Jongmin Lee
Offline reinforcement learning aims to learn a policy solely from fixed datasets, which often contain multimodal action distributions. Flow policies can naturally represent such multimodal behaviors, but learning an efficient one-step flow policy remains challenging: standard value guidance often leads to mode collapse or exploits overestimation bias in out-of-distribution regions. To address this, we introduce One-step Flow policy via Optimal Transport (OptiFlow), a framework for one-step flow policy learning as a structured sample-allocation problem. OptiFlow jointly trains a value-aware reference flow policy and an efficient one-step policy, coupling their action samples through state-wise entropic optimal transport. For each state, critic-estimated values define the priority of distillation target actions, while the action-distance cost ensures geometrically compatible pairings. By avoiding direct critic maximization, our transport-guided approach enables in-distribution exploitation by anchoring the one-step policy to high-value, dataset-supported modes without the risk of out-of-distribution divergence. Experimental results demonstrate that OptiFlow effectively captures optimal multimodal behaviors and achieves strong performance across diverse offline RL benchmarks. Our code is available at https://github.com/Yonsei-DILLab/OptiFlow.
In preparation for Higgs factories and energy-frontier facilities, future colliders are moving toward high-granularity calorimeters to improve reconstruction quality. However, the cost and construction complexity of such detectors is substantial, making software-based approaches like superresolution an attractive complementary tool. This study explores integrating superresolution techniques into an LHC-like reconstruction pipeline to effectively enhance calorimeter granularity and suppress noise. We find that this software preprocessing step significantly improves reconstruction quality at fixed detector granularity, narrowing the performance gap to a finer calorimeter without closing it. To demonstrate its impact, we propose a novel transformer-based particle-flow model that offers improved particle reconstruction quality and interpretability. Our results serve as a proof of concept, demonstrating the potential of superresolution to enhance reconstruction in collider environments.
Wenlin Chen, Mingtian Zhang, Jiajun He, Zijing Ou, José Miguel Hernández-Lobato, Bernhard Schölkopf, David Barber
Score-based distillation methods train one-step diffusion models in two stages: they first train a teacher score model, then distill it into a one-step student model. However, this distillation process can introduce bias from two sources: errors in the teacher score model and student score estimate. We propose DiffRatio, a new approach for training one-step diffusion models without teacher supervision based on a density-ratio formulation that connects score-based distribution matching and GAN-style adversarial training. Instead of using a teacher score model to provide training targets, DiffRatio directly learns a log density ratio between the student and data distributions across diffusion time steps. As a pre-trained score model is only used to initialize the one-step generator and does not supervise training, this design simplifies the training pipeline, mitigates gradient estimation bias, and reduces the size of auxiliary networks. The learned density ratio can be used as a verifier, enabling a principled inference-time parallel scaling method that further improves sample quality without external rewards or extra sequential computation. DiffRatio achieves strong one-step generation results on CIFAR-10 and ImageNet (64x64 and 512x512), outperforming most teacher-supervised distillation methods.
Jan Blechschmidt, Oliver G. Ernst, Moritz Poguntke, Björn Sprungk
Sampling from the posterior is the central task of computational Bayesian inverse problems. The standard workhorse in Bayesian inference---Markov chain Monte Carlo (MCMC)---is sequential, yields correlated samples, and must be rerun for each new observation. Conditional flow matching offers an amortized alternative: a transport map, trained once on joint samples of parameter and data, that yields independent approximate posterior samples for any observation at negligible online cost, without new likelihood evaluations. We give a careful, MCMC-literate assessment of flow matching for PDE-based inverse problems with function-valued parameters. Exploiting the flow's tractable density, we derive computable accuracy estimates for the learned posterior in total variation distance and Kullback--Leibler divergence and, moreover, propose a hybrid sampler that is made asymptotically exact by Metropolization. We validate the accuracy estimates and demonstrate amortization in several numerical examples, including electrical impedance tomography and a likelihood-free Lotka--Volterra model. Flow matching reproduces posterior means and credible bands closely at a fraction of the forward solves required by MCMC, but its accuracy degrades for concentrated posteriors and for observations that are atypical under the prior.
Detecting fake interactions in digital communication platforms remains a challenging and insufficiently addressed problem. These interactions may appear as harmless spam or escalate into sophisticated scam attempts, making it difficult to flag malicious intent early. Traditional detection methods often rely on static anomaly detection techniques that fail to adapt to dynamic conversational shifts. One key limitation is the misinterpretation of benign topic transitions referred to as concept drift as fraudulent behavior, leading to either false alarms or missed threats. We propose a two stage detection framework that first identifies suspicious conversations using a tailored ensemble classification model. To improve the reliability of detection, we incorporate a concept drift analysis step using a One Class Drift Detector (OCDD) to isolate conversational shifts within flagged dialogues. When drift is detected, a large language model (LLM) assesses whether the shift indicates fraudulent manipulation or a legitimate topic change. In cases where no drift is found, the behavior is inferred to be spam like. We validate our framework using a dataset of social engineering chat scenarios and demonstrate its practical advantages in improving both accuracy and interpretability for real time fraud detection. To contextualize the trade offs, we compare our modular approach against a Dual LLM baseline that performs detection and judgment using different language models.
Agent skills extend language-model agents with task-specific procedures, scripts, and references, but the tasks and environments they target continually change. Existing methods improve skills in bounded runs and retain only the final artifact, discarding the decision history that later agents need to interpret prior revisions, evaluations, and rejected alternatives. We introduce SkillHone, a harness for continual agent skill evolution grounded in persistent decision history. SkillHone pairs skill revisions with evaluation-side evidence that supplies practice feedback, recording structured histories of diagnoses, revisions, evidence, and outcomes. Role-separated subagents run candidate skills on practice probes with redacted reporting and propose revisions informed by prior decisions, enabling cross-session refinement without rediscovering past rationale. On deep-research benchmarks, SkillHone runs without a pre-integrated search stack and outperforms the commercially backed deep-research agent by 15.8 points on GAIA and 3.2 points on WebWalkerQA-EN, while also exceeding prior skill-evolution methods. We further deploy SkillHone on internal tool-mediated analysis scenarios, where it improves accuracy by an average of 18.8 points across seven settings.
Saad Naja, Nasser Odtallah, Jake Van-Praag, Sarah May
Trading discipline often fails people. Our study tested whether four AI trading agents followed their strategies and risk rules. Offering traders the evidence and an agent evaluation toolkit to separate luck from following the rules, before giving agents meaningful autonomy or capital. Our study analysed 2,627 closed positions from 12 February to 27 August 2026 across the same asset universe and execution conditions. Most positions were simulated against live prices with a 0.09% round-trip fee; 18% involved real Hyperliquid orders. Each position declared a strategy, stop and target, with a limit of five open positions per agent. The fleet finished net positive after fees. Rule adherence was high across all four agents, including the unprofitable one. Profitability tracked how often agents won relative to their break-even win rates. Institutional-style controls ran by default and contained losses through a 29% market drawdown and recovery, with documented exceptions. The discipline held for all four agents. Agents kept 97.5% of gross loss within declared stops, and typical losses closed at declared levels. Declared risk remained stable after losses and broadly stable across calm and volatile months. Twenty-three of 24 runs respected the position limit; logs identify the exception's size and duration. The same live data produced different results across assigned strategies. The trade-log tests can be rerun from the published trade snapshot; regime tests also require the published Bitcoin prices and regime labels. Section 3.5 specifies the stop-breach tolerance. These trade-level results are not account-level returns or proof of a lasting edge. Our study provides an evaluation toolkit to help traders distinguish rule-following from luck before committing capital. For anyone considering allocating capital to agents to trade, discipline should be the entry test, not the only test.
Cantao Su, Anna Wegmann, Esther Ploeger, Dong Nguyen
Diversity is a critical property of text corpora, with direct implications for training data curation, test set design, and model generalization. However, its measurement remains fragmented and poorly understood. Existing embedding-based diversity measures vary widely in how they are computed and can yield substantially different diversity scores on the same data, yet no systematic comparison of their empirical behaviour exists. We make two contributions toward addressing this gap. First, we present a taxonomy of 21 embedding-based diversity measures, organizing them into four families based on their underlying computations. Second, we introduce EmbDivBench, a benchmark that evaluates these measures along three conceptually distinct components (variety, balance and disparity) on both simulated and natural text data. Although Mean Pairwise Distance is the most commonly used measure, our experiments show that it is outperformed by several alternatives on the three diversity components, with Graph Entropy being the best choice overall. Lastly, we distil these findings into practical recommendations for diversity measure selection.
Reinforcement learning (RL) for foreign exchange (Forex) trading requires realistic execution modeling, interpretable reward design, and action spaces that respect market constraints. Existing approaches often simplify these aspects, limiting policy transparency and practical relevance. This paper presents a modular RL framework integrating three components: (1) a friction-aware execution engine enforcing anti-lookahead semantics---observations at close_t, execution at open_t+1, and mark-to-market at close_t+1---while modeling spread, commission, slippage, rollover financing, and margin-triggered liquidation; (2) a decomposable 11-component reward architecture with fixed, pre-specified weights and per-step logging, enabling attribution, ablation, and auditability; and (3) a 10-action discrete interface supporting scaling, reduction, closure, and reversal, with legal-action masking enforcing margin-aware feasibility during training and evaluation. Empirical evaluation on EURUSD training-split data reveals non-monotonic interactions among reward components, demonstrating that additional penalties do not consistently improve performance. The full reward configuration achieves a terminal training Sharpe ratio of 0.765 and cumulative return of 57.09%. Comparing action spaces, the extended 10-action interface increases cumulative return but reduces Sharpe relative to a conservative 3-action variant, revealing a return-activity trade-off under a fixed training budget. Scaling-enabled strategies consistently reduce drawdowns, with the combined configuration delivering the strongest endpoint performance. Supplementary diagnostics on GBPUSD, USDJPY, and AUDUSD exhibit qualitatively similar trends. These findings demonstrate the framework's diagnostic value and underscore the importance of controlled, interpretable design in RL-based financial systems.
Irregular multivariate time series forecasting (IMTSF) requires capturing temporal dynamics and cross-variable dependencies from sparse, asynchronous observations. Existing methods often rely on learned temporal encoders, coupling historical representation construction with model optimisation and repeatedly processing the same observations. We propose CTLite, a lightweight forecasting framework that decouples historical summarisation from learned prediction. Specifically, a parameter-free continuous-time moment encoder summarises local levels and trends at multiple temporal scales while suppressing poorly supported slope estimates. A compact readout combines damped local extrapolation with a low-rank cross-variable correction that accounts for observation freshness, target availability, and forecast horizon. Since the historical summaries are independent of trainable parameters and forecast queries, they can be reused across optimisation steps and repeated predictions for an unchanged history without altering predictions or parameter gradients. Experiments on four real-world benchmarks show that CTLite achieves the best or second-best mean MSE and MAE among the evaluated methods with only 88-2,537 trainable parameters. Under matched GPU workloads at batch size 32, CTLite reduces inference latency by 18.6-87.3% compared with the latest models, including the cost of historical encoding. Reusing summaries further reduces total runtime by approximately 52% over 16 forecast calls on the same history compared with recomputing the encoder for every call, including the initial encoding cost. The code is at: https://github.com/SSMa913/NICLab-CTLite.
Before a machine learning model can learn a thermodynamic equation of state, it must discover what its measurements represent: which channels scale with system size, which are intensive conjugates, how sectors pair through contact, and which potential governs stability. When sensors expose only an unknown linear mixture of extensive states and intensive responses, passive observations cannot disentangle physical quantities from coordinate artifacts. We formulate the problem of discovering this hidden thermodynamic ontology directly from anonymous controlled experiments. We present an operational identifiability theory and a constructive polynomial-time algorithm that extracts extensive and intensive scaling sectors from replication contrasts, recovers their dual cotangent pairing from thermal contact and reciprocity, verifies a globally admissible concave potential via discrete cyclic concavity, and determines an invariant matroid of reservoir ensembles. We prove that the residual observational equivalence is strictly (x, lambda) ~ (A x, a A^{-T} lambda + beta), establishing the sharp observational limit that no permitted experiment can break. Blind evaluations on van der Waals fluids and Curie-Weiss magnets confirm robust recovery under ill-conditioned mixing, correctly resolving anonymous Maxwell tie-lines while rejecting non-equilibrium continuations. External validation across six real fluids from the NIST WebBook demonstrates that operational ontology discovery transfers across real physical substances without coordinate leakage.
Hicham Eddoubi, Jonas Ricker, Federico Cocchi, Angelo Sotgiu, Maura Pintor, Marcella Cornia, Lorenzo Baraldi, Asja Fischer 等 9 位作者
AI-generated images (AIGIs) have reached a level of quality at which humans are unable to reliably distinguish them from real images. To counter the inherent risks of fraud and disinformation, their detection is a pressing challenge and an active research topic. While many of the presented methods claim high detection accuracy, they are often evaluated without considering their adversarial robustness. In real-world scenarios, malicious actors can modify AIGIs, e.g., through imperceptible perturbations, to evade detection systems. In this work, we provide a dataset to assess the robustness of AI-generated image detectors. We present RAID (Robust evaluation of AI-generated image Detectors), a benchmark dataset of 216k diverse and transferable adversarial images. The proposed dataset is created starting from images generated by four different text-to-image models (to promote diversity) and optimizing attacks against an ensemble of seven state-of-the-art detectors (to promote transferability). Extensive experiments show that our methodology generates adversarial images that transfer to unseen detectors, degrading the performance of every detector outside the attack ensemble that we evaluate, and provides a practical, worst-case robustness screen: failure on RAID is a strong indicator that a detector is not adversarially robust. We release our dataset at https://huggingface.co/datasets/aimagelab/RAID and evaluation code at https://github.com/pralab/RAID.
Accelerated dynamics methods, such as parallel replica dynamics, require estimates of the decorrelation time, after which a process trapped in one of its metastable states is distributed according to the quasi-stationary distribution (QSD). In a method proposed by Binder, Lelièvre and Simpson, decorrelation times are defined by coupling a particle approximation of the non-linear conditioned probability flow of the process with MCMC-inspired diagnostics. We explore an alternative to MCMC diagnostics, based on coarse-graining of non-linear probability flows and simulation-based inference. We first develop a coarse-graining framework for killed Markov jump processes conditioned on survival, proving uniform-in-time, non-asymptotic estimates approximating the non-linear probability flow of the true process by the non-linear flow of a surrogate process, defined with a collective variable. The framework does not assume reversibility of the true process, and also gives bounds which scale explicitly in an adiabatic regime. Motivated by these results, we then construct a quasi-stationarity diagnostic using simulation-based inference: a recurrent neural network (RNN), trained on particle approximations of non-linear effective flows, estimates posterior probabilities of decorrelation. The learned diagnostic is then transferred to collective-variable traces of high-dimensional particle approximations. We investigate the behavior and robustness of this diagnostic on a set of numerical experiments, comparing it to the MCMC diagnostic of Binder et al.\ on both synthetic examples and a molecular system, in and out of its training prior distribution. When its prior assumptions hold, the RNN diagnostic outperforms the MCMC diagnostic in ROC analysis, is less conservative, fails less often, and has an interpretable decision threshold.
Michele Caprio, David Stutz, Shuo Li, Arnaud Doucet
An open question in \emph{Imprecise Probabilistic Machine Learning} is how to empirically derive a credal region (i.e., a closed and convex family of probabilities on the output space) from the available data, without any prior knowledge or assumption. In classification problems, credal regions are a tool that is able to provide provable guarantees under realistic assumptions by characterizing the uncertainty about the distribution of the labels. Building on previous work, we show that credal regions can be directly constructed using conformal methods. This allows us to provide a novel extension of classical conformal prediction to problems with ambiguous ground truth, that is, when the exact labels for given inputs are not exactly known. The resulting construction enjoys desirable practical and theoretical properties: (i) conformal coverage guarantees, (ii) smaller prediction sets (compared to classical conformal prediction regions) and (iii) disentanglement of uncertainty sources (epistemic, aleatoric). We empirically verify our findings on both synthetic and real datasets.
Deep Neural Networks (DNNs) are increasingly being deployed in security-critical and safety-sensitive applications, which makes rigorous testing essential to identify and mitigate model weaknesses. Existing DNN testing approaches explore either the input space or a learned latent space. While latent-space generation can better maintain plausibility than direct input-space mutation, current methods still face a trade-off among exploration controllability, failure diversity, and seed-relative semantic drift. To overcome these limitations, we propose Latte, a black-box testing framework that generates semantically proximate, diverse, and fault-revealing test cases by leveraging the latent space. Specifically, Latte encodes each input seed with a pre-trained VQ-VAE and performs a seed-centered, one-step latent mutation along directions defined by anchors sampled from alternative classes, followed by quantization and decoding back to the input space. This explores local neighborhoods around each seed within the learned latent manifold, resulting in a larger number and broader diversity of oracle-triggering prediction discrepancies under the same budget. We evaluated Latte on 5 datasets and 10 DNN models in single-model and multi-model testing scenarios. Across the evaluated datasets and models, Latte improves fault exposure and behavioral diversity under matched testing budgets. Under the single-model setting, it also maintains low seed-relative semantic drift with respect to the source seeds.