The paper proposes a two-loop inference architecture inspired by System 1/System 2 cognition: a fast loop (MobileNetV3-Large) classifies an image and, if confidence is low, a medium-speed loop (DINOv2-B) processes additional evidence. The authors evaluate this REASON framework on four fine-grained image classification benchmarks under distribution shift and synthetic corruptions, showing gains over the fast loop alone at roughly half the computation of the full medium model.

This paper tackles the practical problem of choosing detectors and trackers for multi-pedestrian tracking across scenes with different densities. It carries out a broad empirical comparison of the YOLO11, RT-DETR, and YOLO26 detector families paired with ByteTrack and BoT-SORT on MOT17, MOT20, and SOMPT22, including an ablation of NMS versus end-to-end inference for YOLO26. The study offers a useful survey of quality-speed trade-offs, although its conclusions are constrained by out-of-the-box COCO weights and very small benchmark slices.

VLX-Go targets embodied visual tracking in distractor-rich shared spaces by casting local motion as vision-language-conditioned short-horizon waypoint prediction. It combines frozen DINOv3 and SigLIP features, multiscale temporal context, and a transformer backbone to predict $K=8$ robot-centric waypoints. A staged training recipe—vision-language alignment, multimodal reasoning mid-training, and tracking specialization—drives a large improvement over the OmTrackVLA baseline, while preliminary DAgger fine-tuning provides modest further gains. This matters because existing methods suffer severe success-rate drops when moving from single-target to distractor tracking.

This PhD thesis tackles novelty detection across weather sensor networks and computer vision. For sensor networks, the core problem is detecting blocked rain gauges that falsely report zero rainfall—a failure mode that is hard to catch because zero is also the most frequent correct reading. The work benchmarks existing quality-control methods and proposes new joint-reasoning algorithms (PCSD-SAT, PCSD-JIGM, PCSD-JIKR) to improve detection of spatially clustered faults. For computer vision, it claims to introduce the NOVEL-SS benchmark and a Detect-Characterize-Accommodate (DCA) baseline for handling multiple simultaneous visual novelties. The research matters because reliable precipitation data underpins agriculture and disaster management in data-scarce regions, while real-world vision systems must cope with diverse, concurrent anomalies rather than single-type novelties.

MMKG-R1 attacks the problem of building useful multimodal knowledge graphs (MMKGs) for downstream Graph RAG. Existing methods either construct graphs without task-aware training or rely on reinforcement learning driven only by final-answer correctness, which can give credit to noisy graphs that happen to retrieve the right image. The core idea is to train a Qwen3-VL-4B constructor with GRPO using three complementary, quality-aware rewards: a graph-level Deducibility Reward and two source-level rewards (Source-Grounded Quality and Graph-Readiness) that directly assess extraction fidelity and structural usability. This matters because a correct downstream answer does not guarantee that the MMKG itself is accurate or reusable across queries.

NavAble tackles a critical data scarcity problem in assistive navigation: standard vision datasets largely ignore accessibility-critical objects such as pedestrian signals and door-activation buttons, leaving blind and low-vision assistive systems with unreliable perception. The authors introduce a large-scale dataset and a synthetic generation pipeline that combine 36K curated real-world images, 5.5K newly collected egocentric frames, and 452K synthetic frames across 11 BLV-relevant classes. By filling this gap, NavAble provides a foundation for training segmentation models that can recognize navigationally important objects from diverse assistive-device viewpoints.

Robotic ultrasound lacks safe, realistic simulators because physics-based phantoms and acoustic solvers exhibit a large domain gap with clinical imaging, while trial-and-error on patients is unsafe. SonoWorld proposes a three-stage pipeline that first pretrains a latent-action video diffusion model on roughly 6.9 million unlabeled clinical frames, then grounds the learned dynamics to 6-DoF probe poses via a small EM-tracked thyroid dataset, and finally distills the model into a causal autoregressive generator with selective KV-cache retention for streaming inference. The goal is to provide a data-driven interactive environment for training and evaluating robotic scanning policies without requiring per-subject pose tracking, although the achieved fidelity and limited evaluation leave its practical readiness unclear.

Equivariant GNNs for molecular modeling are bottlenecked by Clebsch-Gordan tensor products, which suffer from high complexity and limited non-linearity. This paper proposes the Equivariant Spherical Transformer (EST), a plug-and-play module that applies self-attention over sampled points on the spherical spatial domain of group representations, using a Fibonacci Lattice to preserve approximate SO(3) equivariance. The work matters because it reports state-of-the-art results on OC20 and QM9 while empirically breaking the degree-limited expressiveness barrier that constrains traditional tensor-product architectures.

Multi-teacher on-policy distillation (MOPD) fuses specialized RL teachers into one student, but the mechanistic path from supervision choices to parameter changes and task capabilities has been unclear. This paper traces that path through controlled experiments on Qwen3-1.7B and SmolLM3-3B, using identical student parameters, responses, and optimizer states to compare gradients, Adam updates, and learning curves across loss-averaging rules and vocabulary truncations. Its core insight is that stronger gradient approximations do not necessarily yield better task performance; instead, outcomes depend heavily on implicit response weighting and optimizer dynamics that prior recipes treat as secondary.

The paper asks whether effective resistance in tissue-specific protein interaction networks can flag which protein-function predictions a graph neural network will get wrong. Because effective resistance on these networks is almost completely determined by inverse degree, the author strips out that degenerate component and tests whether the residual predicts per-node loss beyond standard uncertainty scores. The study concludes that a reproducible but tiny signal exists across 24 interactomes, and that degree degeneration strongly bounds how useful resistance can ever be on this class of biological graphs.

Ablation studies in language models are confounded by self-repair, where downstream components appear to compensate for a removed part. This paper argues that such compensation is not dynamically recruited but is instead a standing, pre-existing opposition released by the intervention. The authors introduce a continuous dose axis for counterfactual interventions and show that downstream contributions follow an affine law predicted in part by static weights. If the finding generalizes, it would reshape how interpretability researchers infer component importance from ablations.

Intrinsic rewards in reinforcement learning aim to drive exploration by valuing experience, yet maximizing them need not yield the most informative data. This paper proposes comparing policies by counterfactual information: one policy dominates another if its interaction histories can be decoded to reproduce the other's history distributions uniformly across possible worlds. Using a simple alarm environment, the authors prove analytically that count-based, prediction-error, empowerment, and several state-based objectives can have strictly dominated optima, while also characterizing when posterior-based rewards succeed and constructing a native score $J_w$ that preserves every strict exploration improvement.

The paper tackles sampling from Gibbs distributions on matrix spaces with potentials that are neither convex nor globally gradient-Lipschitz. Its core idea is to replace the quadratic kinetic energy of underdamped Langevin dynamics with a non-quadratic, relativistic-like energy $K_\delta(p)=\delta^{-1}\operatorname{Tr}[(I_n+\delta pp^\top)^{1/2}-I_n]$, whose gradient tames momentum singular values smoothly. By carefully coupling a momentum-dependent diffusion matrix $M_\delta(p)$ and its divergence correction $\Gamma_\delta(p)$, the authors preserve the exact target Gibbs measure, recover regularised Muon in the zero-temperature limit, and prove exponential convergence of the diffusion plus uniform-in-time moment bounds for an explicit Euler-Maruyama scheme.

Most LLM agents treat repeated use of the same external source as a sequence of independent retrieval events. This paper reframes the deeper problem as source learning: progressively building a persistent, reusable source model that captures how a source’s knowledge is structured, interpreted, and applied. The proposed SourceLearn system combines self-directed study with task-guided refinement, and the authors report large gains across document, code, API, and interactive benchmarks.

This paper tackles competitive Super Smash Bros. Melee with a single Transformer policy that controls all 26 fighters from one checkpoint. The authors pretrain 10M- and 75M-parameter models on roughly 840k human replays, then apply rank-based supervised curricula, 75M-to-10M distillation, and Fox-only reinforcement learning. The resulting 10M policy wins 240 of 244 benchmark games against prior releases and runs within the 16.67 ms frame budget, while the authors openly acknowledge key confounds in evaluation.

The paper tackles how to teach LLMs new skills after pretraining without erasing old ones. While RL resists catastrophic forgetting by staying on-policy, it struggles when the model cannot generate successful trajectories on its own. The authors flip the problem: instead of changing the training loss, they use an MCMC projection sampler to reshape off-policy expert traces so they look more like samples from the base model while preserving correctness. This lets ordinary supervised finetuning match or beat RL and self-distillation on science, math, and medical tasks.

Robust Markov decision processes with polyhedral uncertainty lack exact convex representations because the natural robust Bellman inequalities become bilinear or union-of-polyhedra constraints. This paper shows how to encode a finite sequence of robust policy-iteration steps into a single exact LP whose optimal face recovers the robust optimal value and all optimal stationary randomized policies. At fixed discount, the LP has polynomial dimension and encoding length, and the authors use this representation together with a unified complexity analysis to prove strongly polynomial bounds for several standard uncertainty sets—including $\ell_1$, weighted $\ell_1$, interval, and Wasserstein—that previously lacked such guarantees.

This paper adapts Non-Linear Multi-Dimensional Concept Discovery (NLMCD) from vision to LLMs to model concepts as low-dimensional manifolds rather than linear directions. It introduces a fuzzy generalized Rand index called Concept-Based Alignment (CBA) that compares independently discovered concept spaces across layers, languages, and training stages without requiring shared tokenizers or architectures. The authors claim CBA is more sensitive than linear baselines (CKA, PCA), uses it to reveal cross-layer block structures, and applies it to study multilingual sharing, cross-model alignment, and post-training drift in Tulu-3.

When a machine-learning benchmark improves, multiple mechanisms could explain why. This paper studies how an autonomous agent should select diagnostic probes from a finite library to discriminate among competing explanations under a tight budget. MECHVAR scores each probe by the posterior-weighted variance of its predicted responses, showing that under a shared Gaussian predictive scale this is exactly proportional to the classical Box–Hill pairwise-KL criterion. The reformulation yields $O(KE)$ vectorized rescoring, an explicit pairwise audit trail, and orders-of-magnitude faster acquisition than quadrature-based expected information gain.

FedMIX-P tackles preconditioner coordination in federated learning by mixing a shared reference preconditioner with each client's local adaptive state at every local step. The core idea is to control the operator mismatch that arises when heterogeneous clients transform gradients differently, and the paper demonstrates empirical improvements on vision and language models under severe data heterogeneity. However, the convergence analysis is restricted to linear positive-definite preconditioners, while the headline experiments rely on nonlinear methods such as Muon, creating a notable gap between theory and practice.