Papers

Autonomous reasoning is among the most scientifically and economically motivating topics in AI today. Historically the purview of symbolic AI, recent advances have mainly emerged from deep probabilistic generative models. Despite immense interest and rapid progress, the generative AI community has not clearly converged on operational definitions for reasoning and often implicitly rejects the historical treatment of this topic in logic and verifiable automated reasoning. This position contends that definitional ambiguity leaves the construct validity of reasoning evaluation unverifiable, undermining quantifiable progress toward trustworthy autonomous reasoning. We also contend that this ambiguity is addressable. To that end, we provide (1) operational definitions based on a synthesis of the literature, positioning valid and sound reasoning as a learnable rule-based process; and (2) a checklist for best practices in the communication of AI reasoning research.

Grading natural-language mathematical proofs is a recurring cost in evaluating math-reasoning systems, and frontier LLM judges are expensive. We ask whether cheap open-weight models can serve as reliable judges given a candidate proof, a ground-truth proof, and a human-grading rubric. On a 200-instance validation sample of IMO-GradingBench, two of three cheap judges (GPT-OSS-120B, DeepSeek-V4-Flash) and their three-model consensus are statistically no worse than the frontier (Claude Opus 4.7, Gemini 3.1 Pro) on agreement with human pass/fail decisions, at 4-100$\times$ lower cost. On the full 1000-instance benchmark, the choice of consensus rule over the three judges is a precision/recall dial: unanimous (all-three-pass) rules reach the highest precision (0.855), majority vote the highest recall (0.912); across four replicate runs the unanimous rule is also the steadiest. No rule won outright; the dial replicated on a held-out 600-instance split and on the independent ProofBench. In this domain, cheap judges are competitive with the frontier at one to two orders of magnitude lower cost, and unanimity is the right setting when false positives are costly.

LLM-as-a-judge has become the standard for automated evaluation, but it suffers from high cost, inference latency, and opaque decisions—limitations that undermine its scalability and reliability. We address these with a simple, efficient alternative: program distillation. Instead of prompting an LLM at evaluation time, we distill its decision logic into a committee of programs that can score candidates directly. These programmatic judges offer transparency, are easily inspected or edited, and eliminate per-sample API costs. Building on this notion, we introduce PAJAMA, a system that synthesizes programs as judges, aggregates their decisions into a joint verdict, and incorporates a fallback mechanism to selectively escalate low-confidence cases to an LLM. Across five datasets and eight model families, we show that programmatic judges match the performance of a 13B-size LLM judge at 47x higher throughput. When using program outputs as routing signals, PAJAMA improves both accuracy and throughput and advances the Pareto frontier. Beyond evaluation, programmatic judges produce cheap and effective reward signals: on RewardBench, a reward model distilled from programs' verdicts outperforms one trained on a proprietary LLM's labels at two orders of magnitude lower API cost.

We introduce Office Comprehension Bench (OCB), the first public benchmark to jointly evaluate LLM systems on Word, Excel, and PowerPoint comprehension over native file formats (.docx, .xlsx, .pptx) and their variants. OCB consists of two tracks. File Fidelity Q&A tests structural and visual perception of office artifacts - tables, charts, embedded images, formulas, and app-specific elements such as headers, speaker notes, and named ranges. Domain Q&A tests expert-level reasoning grounded in real-world industry documents across 12 professional domains, with queries requiring multi-step analysis and synthesis across documents. Each reference answer is decomposed into atomic, binary-gradable claims, and an ensemble of LLM judges scores responses against each claim independently. Even the strongest frontier system in its default reasoning mode reaches only about 59.3% on Domain Q&A; increasing thinking depth within a tier does not move performance materially, while moving to a higher product tier yields modest gains. We release the dataset, evaluation tooling, judge prompt, and a public leaderboard.

CONTEXT: Agent skills package reusable behavioural guidance and resources that large language model agents load or apply during a run. Within the agent harness, agent skill harnessing comprises connected architectural responsibilities that govern which skills enter a run, which actions they can cause, and how their use is recorded and informs revision. OBJECTIVES: We identify architectural patterns for agent skill harnessing and organise their responsibilities into a reference architecture that makes these connections explicit. METHODS: A multivocal literature review covered 36 practitioner systems and 51 academic papers; we consolidated practices into patterns and derived the reference architecture from patterns, source evidence, and architectural reasoning. We evaluated the reference architecture by mapping eight systems and applying it in Deep Agents and Hermes Agent. RESULTS: Ten patterns cover skill supply and activation, skill declarations and execution authority, verification and recording of skill use, and run evidence feeding repair and governed revision. The reference architecture organises their responsibilities into four layers (Supply Chain, Mediation, Execution Control, Evidence & Feedback) and two cross-cutting substrates, separating permission determination from enforcement and routing persistent revisions through acceptance. Mapping eight systems showed the first three layers broadly realised and 14 of 15 partial or unestablished ratings in Evidence & Feedback. Applying the architecture in Deep Agents and Hermes Agent turned requirements into checks that exposed a shell route outside the file-tool rules and a preview–apply mismatch. CONCLUSION: The patterns and reference architecture support comparing implementations and turning requirements into responsibility allocations and checks.

Accurate modeling of quantum many-body systems often requires computationally expensive simulations such as Density Matrix Renormalization Group (DMRG) or Quantum Monte Carlo (QMC) calculations. These methods, while precise, impose significant time and resource constraints, limiting their use in exhaustive parameter exploration. Moreover, these expensive simulations can contain variable errors over the large unknown parameter space, which needs to be quantified and propagated. Thus, predictive modelling is required to estimate the functional space accurately over scarcely sampled data with heteroskedastic noise, while preserving the physical relevance of the estimation. Therefore, we present a Physically Constrained Ensemble Gaussian Process (pc-EGP) framework designed to efficiently model complex and noisy quantum systems under physical consistency constraints. The proposed method first enforces physical constraints as a user controlled weighted penalty to the data-driven loss function of the Gaussian Process (GP) surrogates. Then an ensemble of such GP models is trained with variable noisy simulations via numerical quadrature method where these multiple GP(s) at different nodes is integrated as a quadrature weighted average. We first demonstrate the framework on synthetically generated data before applying to quantum systems. In the first case study, we leverage DMRG simulations of the Bose-Hubbard Model to predict the critical interaction parameter Uc governing the superfluid-to-Mott-insulator transition. In the second case study, we demonstrate our method on QMC simulations, of a quantum liquid confined inside a nanoporous silicate with the goal of optimizing a chemical environment to realize a one-dimensional superfluid. Compared to conventional GP, pc-EGP achieves a better balance of accuracy and physically meaningful predictions.

Training deep spiking neural networks (SNNs) remains challenging due to sharp loss landscapes and temporal inconsistency caused by surrogate gradients. To address these challenges, we propose a unified framework: adaptive and asymmetric surrogate gradients A2SG. The adaptive gradients adjust an effective window for spatio-temporal adaptation, reducing spatial gradient variation and maintaining directional consistency of gradients over time. The asymmetric gradients reflect neuronal dynamics by assigning larger gradients to neurons with higher membrane potentials, and we prove that they yield lower variation than symmetric surrogates. Our analysis further establishes a direct connection between local gradient variation and the curvature of the loss landscape, providing a principled explanation for how A2SG promotes convergence to flatter minima and improves generalization. We conduct extensive experiments on diverse models, including CNN-based and Transformer-based SNNs, across various tasks such as image classification using both static and neuromorphic datasets, as well as segmentation. The results demonstrate that A2SG consistently improves accuracy and energy efficiency, establishing it as a general and reliable solution for training deep SNNs. Our code is available at https://github.com/KIST-NCL/A2SG.git.

A key step in knowledge discovery is the evaluation of data mining results. In several applications, including pattern mining, graph analysis, and others, this step includes the evaluation of the statistical significance of the results, to avoid spurious discoveries due only to noise or random fluctuations in the data. While specialized procedures have been developed for some specific applications, resampling-based approaches are widely used, in particular for complex analyses where analytical results cannot be derived. However, current resampling-based approaches require the generation and analysis of thousands of resampled datasets, and are therefore impractical for large datasets or computationally intensive analyses. In this paper, we introduce FewRS, a simple and effective resampling-based approach to assess the statistical significance of data mining results with rigorous guarantees on the probability of false discoveries. Our approach can be used in every situation where resampling-based approaches are applied. FewRS builds on our derivation of a novel bound to the supremum deviation of test statistics representing the quality of data mining results. We prove that FewRS needs to generate and analyze an extremely small number of resampled datasets, leading to a highly scalable approach with wide applicability. We test our approach on common tasks such as pattern mining and network analysis. In all cases, our approach results in a reduction of up to two orders of magnitude in running time compared to the state of the art, while preserving high statistical power, enabling the statistical validation of data mining results on large-scale real-world datasets.

Scaling laws provide a fundamental framework for understanding the performance of Language Models (LMs), yet deriving them requires prohibitively expensive evaluations across thousands of checkpoints or millions of inference samples. To address this, we introduce Item Response Scaling Laws (IRSL), a unified framework that integrates Item Response Theory (IRT) within the scaling law framework. Unlike traditional approaches that treat each model-benchmark pair in isolation, IRSL disentangles latent model ability from question characteristics, factorizing the scaling law estimation for $M$ models and $N$ questions to significantly reduce parameter complexity from $O(M \times N)$ to $O(M + N)$. We instantiate IRSL with Beta-IRT, which leverages the empirical probability responses of LMs – such as token probabilities in pre-training and pass rates in test-time sampling – to capture richer signals than binary responses. We validate our approach across two prevalent scaling paradigms: (1) pre-training downstream scaling, using 6,612 LM checkpoints and 37,682 questions from 10 benchmarks; and (2) test-time scaling, using 12 LMs and 120 questions from 4 benchmarks with up to 2,500 samples per question. Given a one-time calibration on existing model responses, IRSL yields more reliable scaling estimates using only 50 questions per benchmark (a 99.9% reduction), achieving comparable or superior decision accuracy to traditional approaches. Furthermore, we show that the estimated latent model abilities are generalizable, enabling accurate performance forecasting across benchmarks that share the same measurement objective.

Deep neural networks often contain redundant hidden units. Removing individual weights can reduce parameter count, but unstructured sparsity is not always easy to exploit in standard dense implementations. This paper develops a structured pruning framework in which complete neurons are removed using multi-armed bandit (MAB) algorithms. Each candidate neuron is treated as an arm; pulling an arm temporarily masks that neuron, measures the change in loss on a sampled mini-batch, restores the neuron, and updates an estimate of its safe-removal reward. The framework supports stochastic policies, including Epsilon-Greedy, Softmax, UCB1 and Thompson Sampling, and multiplicative-weight policies, including Hedge-style multiplicative weights and EXP3. We evaluate the method on tabular classification, tabular regression and deep neural-network benchmarks covering image, text and reasoning tasks. Statistical comparisons using the Friedman test followed by the Nemenyi post-hoc test show significant differences between methods. On tabular classification tasks, UCB1 obtains the highest mean rank among pruning policies and improves on the unpruned neural network. On regression tasks, UCB1 obtains the highest mean rank and is statistically competitive with, or superior to, several standard regression models according to R^2. On deep-learning tasks, UCB1 and Thompson Sampling obtain the strongest ranks, and several MAB policies significantly outperform the unpruned model, magnitude-based neuron pruning and greedy activation-variation pruning. The results show that MAB-based neuron pruning is an effective and computationally practical approach for structured model reduction.

Reliable measurement of income and consumption is essential for monitoring poverty and inequality in low- and middle-income countries, yet full household surveys are costly and difficult to implement regularly. This paper examines whether reduced survey instruments can preserve key distributional information. We apply Random Forest Recursive Feature Elimination (RF-RFE) to the 2018/19 Nigeria General Household Survey-Panel to identify the income sources, consumption categories and household characteristics that best classify individuals within the welfare distribution. The analysis focuses on three outcomes: poverty status, location in the quintile distribution and position relative to the Gini-based inequality line. The survey's post-planting and post-harvest periods allow us to assess performance under different seasonal contexts. Results show that RF-RFE achieves strong classification accuracy with few predictors. For consumption, poverty status and inequality-line position are accurately predicted using a small set of expenditure categories, while quintile classification reaches about 80 percent accuracy for seasonal consumption and 60–65 percent for annual consumption predicted from a single seasonal visit. For income, poverty status reaches around 90 percent accuracy with five predictors, and inequality-line position is largely captured by labour earnings. The findings suggest that machine-learning methods can help improve survey design and reduce data requirements while retaining much of the distributional information needed to measure and monitor poverty and inequality.

We argue that many Anthropomorphic Misalignment Research (AMR) studies need stronger evidence to ensure that they can provide a robust foundation for critical safety decisions, such as model deployment and regulation. By evaluating failure modes across different misalignment concepts, such as deception, emergent misalignment, and sycophancy, we show how conceptual ambiguity, non-robust datasets, experimental design, and insufficient causal interventions can lead to overinterpretation of model behaviors. This position paper aims to offer guidance on evidentiary considerations that can help improve methodological rigor in AMR. To achieve this, we provide a clear call to action through a proposed framework of evidence levels and a diagnostic checklist. These shared standards will enable more productive scientific discourse and ensure that claims about AI risks rest on solid empirical foundations.

This paper proposes an improved approach to the analysis of Mining Software Repositories (MSR) datasets via metadata enrichment, FAIRness assessment, and topic-driven analysis. This research expands upon an earlier dataset directory created specifically for the analysis of MSR datasets by adding new annotations to the datasets, enriching the metadata categories, and offering more advanced filtering options. The metadata of the MSR papers presented from 2013 to 2024 has been gathered using the Semantic Scholar API. The analysis is based on Latent Dirichlet Allocation (LDA) topic modeling and statistical analysis. Dataset-level attributes were included into the expanded dataset directory, namely repository hosting site, format, accessibility, reusability, and dataset quality. The study reveals that the choice of repository hosting sites and data formats influences citation patterns and dataset usability. Furthermore, the enhanced annotation approach improves the analysis and discoverability of MSR datasets, supporting more effective reuse and evaluation of research artifacts.

State-of-the-art GRPO-style methods for speech-aware large language model post-training suffer from coarse credit assignment, broadcasting the same terminal-reward advantage to every token in a response. This ignores useful structure within rollout batches, where speech-conditioned completions often share prefixes before diverging at important decisions. We propose Low-rank Exploration with Adaptive Forking (LEAF), a retrospective tree-based RL method that recovers this structure without online branching or additional decoding. LEAF samples complete responses, selects high-surprisal boundaries, groups responses by shared prefixes, and assigns span-level advantages using descendant rewards. We theoretically justify LEAF's span-level credit assignment and boundary-selection design. Empirically, LEAF improves over GRPO across speech question answering and speech translation benchmarks under the same rollout and low-rank adaptation budget. Notably, smaller LEAF-trained models consistently outperform current state-of-art, full-parameter baselines.

We present a systematic study of fine-tuning OpenAI's Whisper large-v3 for Swiss German ASR, using 1,367 hours of broadcast speech paired with Standard German subtitles as weak supervision. Through 16 iterative training runs on an NVIDIA DGX Spark (Grace Blackwell, 128 GB unified memory, up to 1 PFLOP FP4), we compare LoRA and full fine-tuning of the 1.55B-parameter model, investigate hallucination root causes, and quantify the effect of data quality, subtitle alignment, and training strategy. Our best model achieves 25.6% measured WER on the All Swiss German Dialects Test Set (ASGDTS) in an honest evaluation on strictly disjoint data. A harmonized error analysis separating genuine errors from valid stylistic variation (tense, word order, Swiss orthography) yields a content WER (cWER) of 13.8%, counting only actual recognition failures. Bias-corrected estimation reduces this to 8.5%, suggesting the true error rate is roughly one third of measured WER. We demonstrate that published state-of-the-art Swiss German ASR results (17.1-17.5% WER) are inflated by benchmark contamination: a vanilla Whisper model self-trained on the ASGDTS test set with zero Swiss German data achieves 13.88% WER, surpassing all published systems. Experiments with Phi-4-multimodal show an even stronger memorization effect (3.9% WER), revealing that the benchmark primarily measures convention matching rather than dialectal comprehension. We release two models, a LoRA adapter (25.32% WER, 13.9% cWER) and a full fine-tuned model (25.60% WER, 13.8% cWER), among the few publicly available, honestly evaluated Whisper models for Swiss German, under Apache 2.0 with full reproducibility, requiring no institutional data agreements.

Advances in machine learning and computational power have unlocked the predictive potential of the human genome, yet biologists now demand that these models also elucidate the underlying biological mechanisms. While interpretable machine learning (IML) techniques have been increasingly applied to bridge this gap, there has been a pervasive reliance on anecdotal validation: the vast majority of research relies on a single IML method and reports only isolated successful instances. Through a benchmarking study on transcription factor binding, we demonstrate the risks of current practices. We show that different IML methods can often (1) yield contradictory explanations for the same predictions, (2) fail to localize known regulatory motifs, and (3) fail to faithfully reflect the model's internal decision process. In light of this, we argue for a validation framework analogous to clinical trials: just as trials require rigorous design and adverse-event reporting, genomic interpretability must move beyond cherry-picked plausibility toward systematic assessment of consistency, faithfulness, and biological validity. To facilitate this, we propose a tiered framework to guide rigorous evaluation and reporting of genomic IML methods.

Very low birth weight infants (VLBWI) are at high risk of mortality and severe neurodevelopmental impairment, including cerebral palsy, yet reliable discharge-time prognostic stratification remains challenging in high-dimensional and data-limited clinical settings. To address this problem, we propose QDSP, an interpretable structured learning framework that integrates Quota-guided Subspace Sampling (QSS) and Differentiable-decision-guided Structure Perception (DSP). The QSS module constructs stability-aware and low-redundancy feature subspaces through bootstrap-based feature consistency estimation, whereas the DSP module employs differentiable soft oblique decision structures to model nonlinear clinical interactions while preserving traceable decision evidence. The proposed framework was evaluated on a real-world VLBWI cohort comprising 51 infants and further validated on three public medical tabular datasets. On the primary cohort, QDSP achieved an accuracy of 0.9200 and an AUC of 0.9714, outperforming representative machine learning and deep tabular learning baselines, including XGBoost, TabNet, and TabPFN. Across external datasets, QDSP maintained competitive discrimination and calibration under varying sample sizes and clinical distributions. In addition, SHAP-based analyses and differentiable decision-path tracing identified clinically relevant predictors, including cystic periventricular leukomalacia (cPVL) and birth weight, consistent with established neonatal pathophysiological evidence. These results suggest that QDSP provides an interpretable and robust framework for discharge-time risk stratification in VLBWI and may support early individualized clinical decision-making in neonatal intensive care settings.

Fine-grained time series data with high temporal resolution is critical for accurate analytics across a wide range of applications. However, the acquisition of such data is often limited by cost and feasibility. This problem can be tackled by reconstructing high-resolution signals from low-resolution inputs based on specific priors, known as super-resolution. While extensively studied in computer vision, directly transferring image super-resolution techniques to time series is not trivial. To address this challenge at a fundamental level, we propose Super-Resolution for Time series (SRT), a novel framework that reconstructs temporal patterns lost in low-resolution inputs via disentangled rectified flow. SRT decomposes the input into trend and seasonal components, aligns them to the target resolution using an implicit neural representation, and leverages a novel cross-resolution attention mechanism to guide the generation of high-resolution details. We further introduce SRT-large, a scaled-up version with extensive pre-training, which enables strong zero-shot super-resolution capability. Extensive experiments on nine public datasets demonstrate that SRT and SRT-large consistently outperform existing methods across multiple scale factors, showing both robust performance and the effectiveness of each component in our architecture.

Analyzing attention weights has become a standard approach for interpreting the information flow of Large Language Models (LLMs). However, this approach has significant limitations as it neglects the geometric properties of the value vectors being aggregated. To address this gap, we introduce Contribution Weights, a projection-based metric that quantifies a token's influence by accounting for it's attention weight, value magnitude, and directional alignment with the layer output. We demonstrate that contribution weights provide a more faithful measure of token importance, consistently outperforming attention-based metrics in identifying semantically critical tokens across different decoder-only models, tasks, and datasets. Further, our metric enables novel mechanistic analysis of attention sinks. While previous work characterized sinks as passive repositories for excess attention, we reveal they serve an active functional role, suppressing information through a convex relationship between sink rate and output norm, stabilizing representations by opposing the semantic drift of low-confidence tokens.

Large language models (LLMs) exhibit strong reasoning capabilities, yet most LLM-based agents are statically deployed and unable to improve through task interactions. Existing experience-driven methods often rely on memory or heuristics without enhancing the model's ability to learn, treating it as a passive executor and leading to early performance plateaus and limited long-term improvement. To address this issue, we propose MetaEvo, a two-stage framework for continual agent evolution that focuses on improving how the model learns from tasks experience, rather than solely on what it stores. MetaEvo first applies preference-based optimization to enhance the model's ability of principle abstraction, then enables the accumulation and reuse of these principles within a modular agent architecture. Experimental results on diverse reasoning benchmarks demonstrate that MetaEvo consistently outperforms strong baselines, maintains reliable improvement across iterations. These findings validate the effectiveness of meta-optimization in enabling agents to learn from experience and continually enhance their reasoning capabilities.

LLM-based LEGO assembly requires both semantic grounding and physical feasibility. In this paper, we identify a data-induced failure mode, physhack, in which generated assemblies satisfy physical-validity constraints while remaining geometrically misaligned, semantically inconsistent, or poorly calibrated. To address this challenge, we propose a model-based data selection approach that uses only 5% of the original training data while improving semantic and structural alignment across multiple independent evaluators. We further introduce PVPO, a reinforcement learning method that couples physical-validity and voxel-space geometric rewards to strengthen the LEGO synthesis capabilities of LLMs. PVPO-trained DeepSeek-Llama-8B outperforms the frontier VLM Kimi-K3 across semantic, structural, and physical evaluation dimensions, even when Kimi-K3 is provided with a ground-truth image as an additional reference.

We introduce the Laplace-Fourier Neural Operator (LFNO), a unified framework for modeling dynamical systems across transient and steady-state regimes by integrating the spectral advantages of Laplace and Fourier Neural Operators. LFNO employs a dual-branch architecture that explicitly decomposes system dynamics into transient and steady-state components. We evaluate LFNO on nine benchmarks, including three ODE systems (Duffing, Lorenz, and Pendulum) and six PDE systems (Euler-Bernoulli beam, Heat, Reaction-diffusion, Brusselator, Burgers, and Navier-Stokes). LFNO significantly outperforms existing operators on ODE systems, where transient dynamics dominate, and consistently surpasses LNO while achieving performance competitive with FNO on PDE benchmarks. Furthermore, LFNO offers improved stability and physical interpretability through its component-wise decomposition. These results demonstrate that LFNO provides a robust and unified approach for learning complex dynamical systems across multiple temporal scales.

We formulate data propagation through the Transformer, the machine learning architecture powering large language models, as a nonlinear control system on the space of probability measures. For the mean-field Transformer model with self-attention and affine feed-forward layers, we prove that Gaussian distributions remain exactly Gaussian along the induced flow. This invariance reduces the infinite-dimensional measure dynamics to a finite-dimensional bilinear control system governing the evolution of the mean and covariance, reformulates the expressive capacity of Transformers as a reachability problem for prescribed Gaussian moments, and reveals a novel connection with Riccati-type equations from classical filtering and control. For time-varying controls, we prove exact finite-time reachability of any target Gaussian distribution whose covariance matrix has the same rank as the initial one, this rank constraint being an intrinsic invariant of the dynamics. For time-invariant parameters, we derive explicit spectral conditions leading either to asymptotic stability toward positive-definite equilibria or to finite-time blow-up of the covariance. Numerical experiments complement the theory by showing that practical Transformers with Gaussian inputs remain close to moment-matched Gaussian distributions through early and intermediate layers, while Transformers with prescribed attention matrices reproduce the predicted covariance regimes: bounded evolution in stabilizing configurations and blow-up in destabilizing ones.

Ordinal Regression (OR) aims to predict target values with inherent order, underpinning critical applications across diverse domains, from recommender systems to computer vision. Though having evolved from naive regression to discretization-based classification and generation, existing paradigms remain fundamentally constrained by quantization artifacts and the lack of global ordinal topological perception. These methods typically enforce rigid boundary delineations, failing to capture the non-stationary semantic transitions inherent to ordinal data. In this paper, we propose a novel paradigm where OR is formulated as a Continuous Generative Ordinal Regression task. Under the novel paradigm, we introduce DiffOR, a unified framework that leverages diffusion models to recover continuous ordinal values via iterative denoising, thereby enabling the dynamic learning of soft semantic transitions. To explicitly preserve ordinal topology, we devise a Dual-Decoupling Strategy: Spatially, Multi-scale Increment Aggregation decomposes targets into hierarchical continuous increments; Temporally, Dynamic Denoising Perception synchronizes denoising steps with feature frequencies, ensuring robust coarse-to-fine refinement. Theoretically, we show that the proposed method can significantly enhance both representation capability and mechanistic interpretability. Extensive experiments on 12 benchmarks across four domains validate DiffOR's consistent superiority over state-of-the-art methods, establishing a new standard that demonstrates strong potential as a general-purpose solution for universal ordinal regression.

We propose a topological framework for comparing trained Graph Neural Networks (GNNs) by mapping the Stochastic Block Models (SBMs) induced on the graphon-signal space of a Message Passing Neural Network (MPNN) onto the unit $n$-sphere $\sphere^{n-1}\subset\R^n$. The construction rests on three classical pillars: the compactness of the cut-distance graphon space $(\Wo,\cutdist)$ \citep{lovasz2006limits,lovasz2012large}, the Frieze–Kannan weak regularity lemma together with its graphon-signal extension due to \citet{levie2023graphon}, and the Lipschitz continuity of MPNNs with respect to the cut-distance. We show that, for any prescribed tolerance $\varepsilon>0$, a trained MPNN $\Phi$ acting on a sufficiently large graph factors (up to $\varepsilon$) through a step-graphon-signal of bounded complexity, and we construct an explicit measure-preserving map $\Psi_n\colon[0,1]\to\sphere^{n-1}$ that places the SBM regions on disjoint spherical caps. This produces a problem-agnostic, low-dimensional “fingerprint” of a trained GNN that is amenable to visual inspection and to nearest-neighbour search across model zoos, enabling transfer-learning candidate retrieval without retraining. We discuss the obstruction posed by concentration of measure in high dimension – a phenomenon directly relevant to LLM-scale embeddings. We close with five concrete future research directions: hyperbolic and Grassmannian alternatives to the spherical model, Gromov–Wasserstein distances on graphon-signals as an isometry-free alternative to the $n$-sphere map, an information-geometric (Fisher) reformulation of the SBM manifold, persistent-homology fingerprints of layer-wise embedding clouds, and a spectral-distance baseline derived from the graphon eigendecomposition.

Pre-training data mixtures are commonly tuned by running small-scale experiments and extrapolating to the target training budget. When high-quality data is scarce and must be repeated, this extrapolation frequently fails, but the source of the failure has not been isolated. We show that a primary culprit is a repetition mismatch: because high-quality datasets are small, their repetition rate changes as the training budget grows, shifting the optimal mixture in ways that small-scale proxy experiments do not anticipate. A subsampling procedure that matches the target repetition rate controls for this effect. In a two-source setting combining limited high-quality data with web crawl, a single repetition-controlled experiment using only 1/16 of the target tokens recovers a mixture within 0.10 of the optimum on Wiki-Text for a 1.17B parameter model, compared to an error of 0.85 without repetition control. Achieving comparable accuracy without repetition control requires multiple training horizons, consuming 19%, 44%, and 94% of the target token budget when using the results from two, three, and four horizons respectively. With three data sources, the larger mixture space requires more than a single experiment to constrain, but the approach remains effective: at the 757M scale, just two repetition-controlled horizons recover the optimal mixture, outperforming baselines that instead require the full two-source experiments to construct. Our results reveal that repetition dynamics, not scale alone, shape whether small-scale mixture experiments generalize. More broadly, they suggest that data repetition deserves treatment as a first-class variable in mixture optimization, rather than an inconvenient side effect of limited data.

Fine-tuning often introduces spurious correlations alongside task knowledge, causing systematic failures on underrepresented groups. Existing mitigations require retraining, group labels, or curated counterfactual data. We show a simple post-hoc intervention reduces shortcut reliance without any of these: truncating the tail of the SVD of $\Delta W = W_\mathrm{ft} - W_\mathrm{base}$ reduces the spurious-group gap while preserving task accuracy. Across three instruction-tuned models ($0.5$B–$7$B) and four classification benchmarks, top-$k$ truncation reduces the gap on every cell at $<2$ pp accuracy loss, by up to $5\times$ on CivilComments. We propose this works because the shortcut response sits in the tail of the singular ordering of $\Delta W$, a claim about how truncation behaves rather than about the raw singular values, which are broadly distributed and look the same across all four datasets. A controlled boundary case in which fine-tuning has only a shortcut to learn shows the predicted FT-to-base collapse, and bottom-/random-$k$ and matched-rank LoRA controls rule out generic low-rank approximation and rank-constrained training as the explanation. We read this as preliminary evidence that the singular basis of $\Delta W$ is a useful coordinate system for studying what fine-tuning has learned.

Recent analyses of multi-pathway Deep Linear Networks use Gradient Flow to predict a "winner-takes-all" specialization in which path symmetry breaks and each feature concentrates in a single pathway. In this work, we show that discrete Gradient Descent (GD) with a large step size tells a different story. We prove that single-path solutions are sharp minima, whereas distributing signals across pathways reduces sharpness by a factor that decreases with both the number of pathways and depth. Consequently, while early training reproduces the depth-driven symmetry breaking predicted by GF, oscillations at the Edge of Stability subsequently override this tendency and drive the network into a re-balancing phase, where signals redistribute across pathways. Together, these results clarify how depth shapes pathway competition and explain why large-step GD favors shared representations rather than persistent single-pathway dominance.

Two-sample testing is a fundamental tool for detecting distributional differences across scientific domains, but classical tests (including kernel-based tests) can be ineffective on high-dimensional structured data such as images. Recent deep two-sample tests improve sensitivity in these settings by learning informative representations, yet they provide limited insight into which data features drive rejection of the null hypothesis $H_0$. To address this issue, we propose a counterfactual explanation framework for deep two-sample testing that generates sample-level edits moving observations from a source group toward a target group while explicitly reducing the discrepancy measured by the test. Our method combines a diffusion autoencoder with a pretrained deep two-sample test model and optimizes a maximum mean discrepancy (MMD) objective in the test model's representation space to produce plausible counterfactuals. We quantify distribution-level effects through changes in the test statistic and the resulting two-sample p-values. We evaluate the method on synthetic 2D shape datasets and two MRI cohorts. Across both settings, the counterfactual transformations consistently increase p-values relative to the original samples, indicating that the edited source set becomes statistically closer to the target distribution under the test. We measure minimality using LPIPS to ensure the counterfactuals remain close to the original samples. The resulting edits provide interpretable evidence of the features associated with the detected group differences. On MRI, the localized changes are consistent with known anatomical differences between cohorts.

Large Language Models exhibit paradoxical fragility in fundamental arithmetic, implying a disconnect between internal computation and discrete output. By analyzing the residual stream geometry during multi-operand addition, we identify the Iso-Raw-Sum Trajectory (IRST), a geometric structure where representations are anchored by semantic digits and modulated by continuous carry fibers. We propose the Noisy Quantization Model to explain this geometry, framing arithmetic errors as Geometric Slippages caused by internal neural noise pushing a continuous, latent Carry Potential across quantization thresholds. This geometric framework further elucidates Probe Versatility, explaining how lightweight probes can disentangle coexisting latent signals (such as ground truth versus hallucination) from a single activation vector. Finally, we validate these insights through a geometric consistency check method that effectively detects and corrects these quantization failures during inference. Our code is available at https://github.com/RL-MIND/Shape-of-Addition.