The paper studies how classification accuracy evolves when neural networks are trained from very small initialization. It identifies the "accuracy well": a striking non-monotone trajectory where accuracy first rises rapidly, then drops significantly, and only recovers much later. The authors argue this reveals learning stages invisible to the training loss, and they support the claim with experiments across architectures and datasets, plus rigorous theory for deep linear networks.

Prostate MRI segmentation suffers from regional boundary ambiguity, causing standard global-overlap losses to overlook clinically significant surface deviations. This paper proposes Contrastive Uncertainty Learning (CUL), which treats predictive uncertainty as a representation-level supervisory signal rather than a post-hoc filter: low-uncertainty same-class voxels are clustered in embedding space while high-uncertainty or misclassified voxels are actively repelled. Evaluated on 589 lower-pelvic MR scans across eight anatomical structures, CUL preserves mean Dice while significantly improving surface-sensitive metrics and suppressing extreme errors in hard cases, suggesting practical value for radiotherapy planning and contour-acceptance workflows.

Clinical assessments collected across heterogeneous systems create mixed longitudinal records that resist standard total-score linking. This paper proposes Bidirectional Feature Harmonization Decomposition (BFHD), a method that identifies which individual measurements can be reliably translated between two assessment systems in both directions while explicitly separating out non-translatable components. By operating directly in original measurement spaces via a shared invertible backbone and feature-wise output gating, BFHD aims to provide an interpretable decomposition for cross-protocol integration and longitudinal analysis.

Graph collaborative filtering models collapse in strict cold-start settings because zero-degree items sever the message-passing chain. DualGCN tackles this with an asymmetric two-branch distillation framework: a teacher network anchors item representations with frozen semantic priors smoothed over the graph topology, and a lightweight student learns to project unseen semantic features into that collaborative space via cosine distillation and residual mapping. If the claims hold up, it offers a much leaner alternative to heavy multi-modal fusion modules for industrial retrieval.

Latent world models based on Joint-Embedding Predictive Architectures (JEPA) offer a scalable route to embodied planning, but existing methods suffer from representational collapse, heavy compute, and weak long-horizon reasoning. This paper proposes MemWM, a lightweight JEPA world model augmented with a content-addressable associative memory bank that supplies high-level subgoals via Adaptive Layer Normalization. By operating at multiple temporal scales within a shared latent space and enforcing slot diversity through complementary losses, the authors aim to boost long-horizon performance without relying on pretrained encoders or doubling model size. Across four control tasks, MemWM achieves an average success rate of 89.9% with only 19M parameters, outperforming heavier baselines such as DINO-WM in aggregate.

This paper resolves the open problem of optimal dimension dependence for dynamic regret in exp-concave online learning. The authors propose SparseGauss, a proper full-information learner that achieves a $\widetilde{O}(d^{1/3}T^{1/3}P_T^{2/3})$ rate via sparse Gaussian histories and square-root KL movement control. A matching lower bound on a Euclidean ball establishes minimax optimality, up to logarithms, when $T \gtrsim d^2$.

Score-based generative models with heavy-tailed $\alpha$-stable Lévy noise have shown empirical promise for extreme-event modeling, yet theoretical guarantees remain scarce. This paper develops a non-asymptotic $W_p$ convergence framework for scale-preserving $\alpha$-stable dynamics and proves that global Wasserstein convergence alone does not separate stable from Brownian models on heavy-tailed targets. A sharp distinction emerges instead in the tails: stable dynamics inherit power-law tails with index $\alpha$, while Brownian outputs stay light-tailed, and asymptotically correct extreme-quantile scaling requires matching the target tail index $\nu$ to the noise index $\alpha$. These results establish tail-index matching as a principled design mechanism for capturing heavy-tailed phenomena.

LLM unlearning is usually tested by whether a model refuses harmful queries and retains clean-task accuracy. The authors highlight a gap: models can pass those checks while failing when benign prompts sit just outside the forgotten domain, a failure they term forget-adjacent utility loss. They propose RA2, which redirects forget-token hidden states toward nearby retain-supported anchors instead of simply pushing them away. Experiments on WMDP and MUSE show this directional choice meaningfully reduces neighborhood-level collateral damage.

This paper investigates whether high-performing humanoid control networks can be synthesized directly in weight space rather than trained conventionally. The authors adopt a latent diffusion framework---pairing a cluster-wise linear autoencoder with a Diffusion Transformer---to generate weights for 3-layer MLP policies on a Unitree G1 robot, achieving zero-shot sim-to-real transfer for walking, dancing, and spin kick. Alongside generation, they analyze the learned weights and identify a kinematic prior in first-layer parameters that emerges under both reinforcement and imitation learning, suggesting embodiment-specific structure in policy weights.

This paper tackles digital-domain self-interference cancellation in in-band full-duplex radios, where conventional adaptive filters degrade when the desired signal is much stronger than the residual interference.
The authors reformulate cancellation as an operator-learning problem: a Transformer neural operator learns the temporal evolution of the self-interference channel from historical observations and reconstructs the current interference from the known transmit signal.
Trained in a desired-signal-free setting and applied during contaminated reception, the method reduces reliance on corrupted instantaneous updates and reports strong high-SIR stability.

Neural symbolic regression models are bottlenecked by training datasets riddled with structural redundancy and simplicity bias. This paper proposes ExCaUS, a deterministic grammar-based framework that exhaustively enumerates canonical univariate skeletons under explicit unary-count and depth bounds, eliminating duplicate functional forms. The authors pair this with a derivative-driven sampling strategy and a 150-problem benchmark, demonstrating that curated structural diversity exposes severe weaknesses in state-of-the-art SR methods.

The paper investigates when jointly optimizing imputation with a downstream classifier improves robustness under train–test mismatches in missingness rates or mechanisms. Instead of treating imputation as fixed preprocessing, the authors cast missing values as learnable variables regularized around a reference imputation, giving both amortized and latent instantiations. Through 98 tabular classification datasets—and targeted shift experiments on 36 datasets—they show that reconstruction quality is a poor proxy for predictive utility and that task-aware corrections matter most in hard regimes such as MNAR and severe distribution shift.

Most standard diversity measures violate at least one of the Mironov–Prokhorenkova axioms (monotonicity, uniqueness, continuity), and the only known measures that satisfy all three are NP-hard to compute. This paper introduces Geometric Kernel Diversity (GKD), which defines diversity as the volume of the union of radial-kernel under-graph solids for a point set in Euclidean space. For a canonical Student-$t_2$ kernel, the authors prove that GKD satisfies all three axioms on Euclidean-realizable inputs, give exact polynomial-time algorithms for dimensions $d \leq 3$, and provide a randomized $(1\pm\varepsilon)$-approximation for arbitrary dimensions. The contribution is a theoretically clean first positive answer in the restricted Euclidean setting, though the axiomatic question on general distance matrices and exact computation in $d \geq 4$ remain open.

Conflicting studies on whether neural networks converge to a shared representation have used different alignment metrics and reached opposite conclusions. This paper proposes the Lattice Representation Hypothesis (LRH), which formalizes alignment criteria into a lattice ordered by strength, and argues that disagreements arise because prior work evaluates criteria at different points on this lattice. It also proves in idealized models that the training procedure and objective determine which criteria are met reliably across independent runs, supported by experiments on dozens of pretrained vision and language models.

GRPO for long-horizon agents suffers from sparse terminal rewards that yield all-fail rollout groups and zero advantage signals early in training. This paper studies guide-assisted curricula that replay successful reference prefixes before the learner continues, systematically varying the initial position, handoff granularity, and movement rule. Across Qwen3 models from 1.7B to 8B on BFCL and AppWorld, the authors find that starting near task completion and using outcome-informed call-level selection improves independent execution while reducing training cost. The work usefully frames curriculum design as a practical lever for GRPO-based agent post-training.

This paper studies noiseless efficient global optimization via Gaussian Process Thompson Sampling (GP-TS) in the frequentist RKHS setting. It proves that standard GP-TS suffers an intrinsic $\sqrt{\log N}$ simple-regret penalty for Mat\'ern kernels, yielding $O(N^{-\nu/d}\sqrt{\log N})$, and proposes Capped Thompson Sampling (GP-CTS), which achieves the minimax-optimal $\Theta(N^{-\nu/d})$ rate using a known RKHS norm bound. The analysis covers both discrete and continuous domains and is notable for matching upper and lower bounds in a non-bandit setting.

RM-OPD targets a real bottleneck in generalist reward modeling: training on mixed-domain data with standard RL produces cross-domain interference and sparse outcome rewards, so capabilities like math verification, code review, and safety guardrails cannot coexist without degrading each other. The core idea is a two-stage framework—first cultivate domain-specialist reward teachers via single-domain GRPO, then distill them into one student through routed on-policy distillation, domain contribution reweighting, and criteria context distillation. This matters because it swaps coarse verdict-level supervision for dense token-level guidance, raising the multi-domain performance ceiling for generative judges.

Autoregressive transformers trained on limit order book streams are usually evaluated only by output realism, leaving their internal mechanics opaque. This paper inverts the problem by training GPT-style models on synthetic order-flow simulators with known latent parameters, enabling exact verification against Bayes-optimal predictors and oracle entropy floors. It finds that next-token pretraining implements amortized Bayesian inference and online filtering, that latent states are linearly decodable from frozen activations, and that regime diversity—not model scale or token count—drives generalization.

Multimodal large language models suffer large accuracy drops when their visual inputs are corrupted by noise, blur, or weather artifacts. This paper asks whether test-time feature alignment—matching activation statistics to clean source moments, a technique that works well for CLIP-like classifiers—can recover that lost performance without labels or weight updates. Across three modern MLLMs and 15 standard corruptions, the authors find it cannot: sample-blind, position-shared corrections recover only about one percentage point of an order-of-magnitude larger drop. The core insight is that corruption shifts in MLLM visual token sequences are nearly orthogonal across images and dominate the residual term in a per-token squared-shift decomposition, whereas pooled classification embeddings exhibit a shared, low-rank shift direction. This matters because it establishes that robustness in generative MLLMs is a per-token restoration problem, not a simple domain-statistics-matching problem.

Latent world models for control are routinely selected and checkpointed by one-step next-embedding prediction loss, on the assumption that lower loss implies better planning. This paper tests that proxy in JEPA architectures by introducing a frozen-decoder protocol that reports one-step error in interpretable physical units, then applies fine-tuning interventions and measures goal-conditioned planning with paired seeds. It finds the relationship saturates after a single epoch within a task and has no stable exchange rate across tasks, implying that latent prediction loss is an unreliable stand-in for planning competence.