Training Efficiency & Optimization - Weekly Roundup

Findings suggest that optimal, high-capacity memory representations are not formed in flat minima, but are dynamically sculpted at the highly curved boundaries of geometric singularities.

Diagnostics limit explanations based solely on terminal optimization accuracy and distinguish numerical cost from quality, rank coverage and freshness; they establish neither practical-tracker convergence nor predictive or deployment benefits from basis availability.

This work proposes a pure single-loop damped extragradient method with fixed parameters and two new full-gradient evaluations per iteration after one initialization query that attains the optimal condition-number order up to logarithmic factors through fixed explicit updates.

A new SF-GVF with prescribed physical speed (PPS) that converge exponentially to the desired path from any initial condition in the higher-dimensional space (including virtual dimension); more importantly, the robot's physical speed converges to the PPS, while the path-error dynamics remain invariant under regular reparameterizations of the desired path.

Diagnostics limit explanations based solely on terminal optimization accuracy and distinguish numerical cost from quality, rank coverage and freshness; they establish neither practical-tracker convergence nor predictive or deployment benefits from basis availability.

This work proposes a pure single-loop damped extragradient method with fixed parameters and two new full-gradient evaluations per iteration after one initialization query that attains the optimal condition-number order up to logarithmic factors through fixed explicit updates.

A new SF-GVF with prescribed physical speed (PPS) that converge exponentially to the desired path from any initial condition in the higher-dimensional space (including virtual dimension); more importantly, the robot's physical speed converges to the PPS, while the path-error dynamics remain invariant under regular reparameterizations of the desired path.

Results indicate that recursive quantum feature transformations can improve stability and out-of-sample performance for compact hybrid quantum--classical temporal models.

A global ambiguity set is constructed as the union of admissible mixtures of local ambiguity sets that allows client-specific ambiguity radii and admits a client-wise separable reformulation and establishes a high-probability out-of-sample performance guarantee.

This work proposes QUALS, a large-scale time series corpus equilibrium framework that significantly enhances data efficiency, enabling existing models to achieve superior performance using only a small fraction of the original training data.

The model presented here is a step toward a more rigorous formalization of cognitive phenomena and may provide a basis for further theoretical analysis, computer simulation, and implementation in artificial-intelligence systems inspired by biological processes.

Inspired by the efficient communication principles of federated learning (FL), this work introduces two hybrid algorithms - FL+FSDP and FL+HSDP - interleaving sharded DP with FedAvg-style aggregations, demonstrating superior computational efficiency and improved model quality.

VideoResearcher achieves state-of-the-art performance among self-improving agents and approaches the human-designed upper bound, demonstrating a training-free paradigm for long-video understanding that expands agent capabilities through autonomous tool development while reducing costly manual engineering.

Dense representation and optimization, not finer sequence modeling, drive CVR AUC at this scale, and verdicts must come from the held-out leaderboard.

SELF-INDEX is proposed, a framework that enables an index to self-evolve without human intervention, and its Optimizer autonomously diagnoses retrieval shortfalls, selectively revises the responsible index keys, and validates each revision before updating the index.

It is shown that steering creates a predictable prompt-injection attack surface whose vulnerability increases monotonically with steering magnitude, relevant to agent systems deploying steered classifiers.

The architecture, training methodology and inference speedups of Granite 5.0 Turbo CTC result in a model that is on the speed-accuracy Pareto frontier of the Open ASR leaderboard for English short-form ASR while being twice as fast as the fastest competitor.

It is demonstrated that LLM annotation reproducibility can coexist with poor construct correspondence and motivate validation procedures that distinguish reproducibility, construct validity, and communication context variation when LLMs are used as measurement instruments.

Group Relative Return Policy Optimization (GR2PO), a critic-free reinforcement learning framework for continuous robot control, is proposed and significantly outperforms critic-free baselines that use immediate rewards and performs competitively against state-of-the-art actor-critic methods.

P-GADMM forms computation-aware edge groups according to client computational capabilities and local data sizes, which reduces training-speed variation within each group, and establishes convergence guarantees for an idealized form of P-GADMM under bounded group-level staleness.

This paper proposes and analyzes Comparison-based Preference Optimization (ComPO), a zeroth-order alignment method based on comparison oracles, and establishes a convergence guarantee for its basic offline scheme under smoothness, gradient sparsity, and compatibility between the oracle and a latent objective.

This work connects Adam loss spikes to both the mismatch between momentum timescales and finite-scale superquadratic loss geometry beyond the Hessian, and shows that confident cross-entropy losses develop a core--wall landscape comprising a narrow quadratic core followed by a steep wall, which produces effective superquadratic behavior at the scale of an optimizer update.

Fuid Notarization is introduced, a notarization paradigm in which document evolution itself becomes the object of notarization that combines two complementary guarantees: deterministic convergence provided by the CRDT and independently auditable proof-of-existence, provenance, and publication evidence provided by the blockchain.

This work presents OmniRisk, an omnidirectional planning framework that learns trajectory-level risk offline for efficient onboard evasion and enables efficient onboard avoidance, with real-world flights demonstrating consecutive evasive maneuvers at relative encounter speeds up to 15 m/s without fine-tuning.

Simulation and real-world experiments demonstrate improved optimization reliability, trajectory consistency, and grasping performance through a CEM-based process that evaluates sampled initializations with the deployed optimizer and retains favorable ones as new supervision.

AccelPact, a parallel runtime enabling zero-I/O in-memory fault recovery for sharded distributed training, and operates directly on native communicator instances, providing an efficient foundation for resilient deep learning.

Two machine-learned representations are compared, DeepTICA and SPIB-VAE, with linear TICA for recovering steady-state observables from flawed distributions and a kinetic score computed from a coarse MSM at a resolution comparable to that used for RiteWeight random clustering is provided.

It is shown that this gating mechanism causes SSMs to first learn an in-weights"memorization"solution, while delaying, or even preventing, convergence to a correct in-context learning solution, and that gating is often beneficial for improving generalization to long sequence lengths.

Overall, FedAvg provided the most consistently reliable server-based baseline, topology-aware gossip offered a viable decentralized alternative, and Hybrid Gossip-FedAvg provided a balanced compromise between peer-to-peer diffusion and periodic global coordination.

Findings suggest that optimal, high-capacity memory representations are not formed in flat minima, but are dynamically sculpted at the highly curved boundaries of geometric singularities.

It is shown that the minimum aggregation is essential: mean aggregation can be exploited, saturating the proxy reward while degrading accuracy, and the minimum aggregation is essential for PROSE, Process Reward Guided Self-Training.

HetShield is proposed, a lightweight trust layer that validates the spatiotemporal consistency across features, recovering 83--95% of the accuracy degraded by attacks, outperforming prior art.

CALOS (Control-Affine Lyapunov On-manifold Safety), a runtime safety layer that enforces attitude constraints on a quadrotor without modifying the underlying learning algorithm, accelerates training convergence and improves data efficiency without producing suboptimal policies.

SWB-DM's cache carries a real warm-up cost, but extending all baselines to the same round budget shows its CIFAR-10 gains are disproportionately large, and on CIFAR-100, FLTrust benefits more -- for reasons entirely unrelated to caching.

The contribution is a bounded application of established optimization to selecting assurance context for a software change; discovering the obligations and downstream agent benefit remain open.

This work presents a highly parallelizable, matrix-free, real-space method for computing the random phase approximation (RPA) correlation energy within Kohn-Sham density functional theory that avoids the explicit construction and eigen-decomposition of the response function matrix and uses Lanczos quadrature to evaluate the trace of a matrix function on a real-space grid.

SongCraft, a latent flow matching based model trained for both generation and fine-grained editing, is presented, showing that a single model can support editing of lyrics, vocal melody, beats, and singer identity and the trade-off between reconstruction quality and editability is studied.

A self-evolving memory paradigm for generative recommendation is proposed, aiming to enable effective evolution across heterogeneous behavioral patterns, and three key principles for effective self-evolving recommendation systems are identified, including isolated memorization, reinforced evolution, and scalable application.

A transferable residual adapter that injects additional ranking-specific features into ranking in a residual manner and an asymmetric multi-epoch training strategy that resets sparse parameters while continuously accumulating dense parameters across epochs, alleviating the overfitting of sparse parameters are proposed.

Four parallelism plans are left unbounded by the parallelism plans in common use, and each grows differently: expert dispatch with the routing matrix, the vocabulary projection with tokens times vocabulary, gradient checkpoint boundaries with depth times sequence length, and optimizer state with parameter count.

This paper reports experiments across six frontier model types from OpenAI, Anthropic, xAI, and Google DeepMind, and uses the term epistemic jailbreak for the accompanying loss of discipline in technical provenance as requested specificity increases.