Nothing here yet
Many Bayesian inference problems involve implicit priors known only through samples, such as generative model outputs or historical ensembles. This paper proposes performing posterior sampling in the Gaussian source space of an improved MeanFlow (iMF) transport map, using parallel tempering with pCN or hybrid split-HMC updates. Its central theoretical advance is proving Wasserstein stability bounds for the implemented few-step posterior in terms of the joint population training loss, removing the need for vanishing ODE discretization error.
Tabular foundation models like TabPFN achieve strong average accuracy but crash on minority groups when subpopulation proportions shift between training and deployment. DR-TFM addresses this by training only a tiny query-scaling network—updating as little as 0.016% of TabPFN-3’s parameters—using distributionally robust objectives on *estimated* groups, with no true group labels needed. Across tabular benchmarks and multiple TFMs, it delivers large worst-group accuracy gains while staying computationally cheap.
Machine learning models often fail under subpopulation shift when spurious correlations mislead them into relying on shortcut features. This paper introduces POTER, a method that derives per-sample importance weights from optimal-transport (OT) dual potentials between the training distribution and a small validation-based reference distribution. By avoiding loss-based proxy selection entirely, POTER aims to remain robust under label noise while requiring only a single standard ERM training stage. The authors report state-of-the-art worst-group accuracy on several benchmarks and identify a previously overlooked failure mode: subgroup-concentrated label noise.
This paper asks whether the best policy from an RL training run is necessarily one of its visited checkpoints. It introduces policy-space scaling—the idea that completed RL histories contain stronger, constructible policies—and proposes SURGE, which fuses two checkpoints via delta-first spectral decomposition without additional training or per-query inference cost. Across two 1.5B math trajectories and one 7B code trajectory, SURGE exceeds the best native measured checkpoint while using fewer reasoning tokens.
Knowledge distillation risks propagating the reversal curse when a teacher can recall a relation forward but not backward. This paper proposes directional label distillation: frozen teachers score candidate answers in the direction they know, and the highest-scoring candidate becomes the student's training target. On parent-child facts from Wikidata, this yields 13-15 point gains in student open-ended accuracy and shows that verification can provide accurate supervision where generation fails.
ReHoPER is a zero-shot, inference-only reasoning method that improves LLM compositional reasoning by formulating it as receding-horizon planning. At each step, it generates a short horizon of candidate intermediate questions, answers one, updates its history, and replans across multiple stochastic paths before aggregating by majority vote. The paper reports consistent gains over self-consistency on MuSR, MoreHopQA, and a new controlled benchmark iLLC, with the largest improvements on the hardest compositional tasks. While the method is task-agnostic, its accuracy gains come at a significant inference-time cost.
Time-series anomaly detection is hard to generalize across datasets because heterogeneous dynamics favor different anomaly criteria. TS-Router solves this by separating transferable temporal representation—learned by a frozen foundation encoder—from target-specific anomaly decisions coordinated by a lightweight router. The router estimates specialist competence from pretrained representations and selects a small subset of detectors to fit unsupervisedly on each target, requiring no target labels at deployment.
ActiveSaddler addresses the overlooked problem of which training scenarios to use during automated LLM-agent harness optimization. It treats curriculum selection as a non-stationary bandit whose arms are dynamically instantiated failure patterns, allowing the scenario distribution to co-evolve with the harness rather than being fixed upfront. On GAIA2 and Terminal-Bench 2.0, this yields test Pass@1 gains of 4.4 and 7.5 percentage points over the same optimizer with a fixed scenario order, arguing that adaptive curricula are a critical new dimension for harness optimization.
Causal representation learning typically requires knowing the causal graph upfront, yet misspecifying this structure discards predictive information and ruins OOD generalization. This paper proposes SaCRL, a kernel-based framework that jointly identifies which of three canonical causal structures governs the data and learns the corresponding invariant representation via HSIC violation metrics and SoftMin aggregation. The key insight is that correct invariances can be satisfied without sacrificing prediction, whereas wrong ones cannot, enabling automatic structure discovery during training.
GUI-HARVEST automatically improves the executable runtime harness around a frozen GUI agent without retraining the model. It grounds diagnosis in before-and-after screenshots from repeated task executions, clusters failures into cross-task behavior modes, and validates candidate code edits against both aggregate score gains and predicted behavioral changes. Experiments on OSWorld-Verified show held-out gains across six diverse backbones, with the largest improvement reaching +12.33 percentage points for Qwen3-VL-32B-Instruct. The work is significant because it demonstrates that GUI-specific multimodal evidence yields more generalizable harness edits than generic text-trace optimizers.
Autoregressive vision-language-action models compress continuous robot motions into discrete tokens, yet deterministic tokenizers like FAST fix a single segmentation even when multiple equivalent token sequences decode to the identical physical motion. TOAST treats action tokenization as a stochastic process, sampling alternative valid segmentations of the same quantized action chunk during training to diversify supervision without collecting new demonstrations. The result is a simple plug-in improvement that yields consistent gains on LIBERO and real-robot tasks, with the largest benefits appearing when training data are scarce.
The paper asks whether reinforcement learning on a small set of expert-built coding tasks can teach a generalizable ability to finish agentic work, rather than merely overfitting to a benchmark format. The authors post-train Kimi K2.7 Code with one epoch of GSPO on a rank-32 LoRA adapter using 1,700 tasks graded by hidden tests and pass-to-pass regressions, then measure transfer to six external benchmarks spanning three agent harnesses, multiple domains, and horizons up to multi-hour builds. The results matter because they probe whether RLVR on narrow, high-quality data yields broad agentic competence or just domain tuning.
Multi-token prediction (MTP) heads are usually baked into the backbone during pretraining on trillions of tokens. This paper asks whether a cheap post-training pass on a frozen model can match that quality. The authors train $K=3$ MTP heads on only $\approx$2.5B tokens of self-distilled chain-of-thought from Qwen3-8B, showing they reach or exceed the mean accepted length (MAL) of jointly-pretrained MiMo-7B. They also propose a chain-aware verification relaxation that drafts faster by tolerating bounded distributional drift, and an online bandit controller that adapts draft depth to batch size at serving time.
Industrial tabular data often suffers from scarcity, class imbalance, and covariate shift between training and deployment. This paper defines the Augmented and Weighted Learning under Covariate Shift (AWL-CS) problem and proposes IGDPR, a framework that steers latent diffusion with invariant potentials and aggregates sample reliability through prototype-based reweighting. The core idea is to synthesize data that aligns with stable decision boundaries rather than outdated source correlations, while filtering out validation noise via structural clusters.
Post-hoc OOD detectors routinely tune hyperparameters or fit combining rules on held-out samples from the same OOD datasets used for final evaluation. This paper shows that the standard validation protocol only prevents memorization of individual images, not memorization of dataset identity, and that most reported gains vanish when the detector is instead evaluated on truly unseen OOD datasets. The authors isolate the effect with two elegant controls and derive a closed-form expression for the inflation.
The paper solves the problem that weight-space task arithmetic is trapped in one architecture because parameter updates share no common coordinate system. Its core idea is the platonic task vector—a model-agnostic object beneath per-model task vectors—and the Universal Task Descriptor (UTD), an architecture-independent matrix of fine-tuning-induced affinity shifts that can be added or negated directly. A closed-form operator folds into the target's last layer, while a low-rank adapter fits the same objective across all layers. The work matters because it enables label-free, weight-free composition of task knowledge across six heterogeneous vision–language and audio–text families.
End-to-end autonomous driving maps raw sensors directly to control actions, yet its real-world bottleneck lies in the training ecosystem rather than model architecture alone. This survey introduces a Data-Strategy-Platform taxonomy that frames training as an interdependent system: data defines what is learnable, strategies govern how learning aligns with driving objectives, and platforms enable scalability. The authors argue the field must shift from raw data volume to value density, from isolated task-specific learners to universal foundation models, and from static offline training to closed-loop training-testing integration.
EyeTAG tackles temporal jitter in video gaze estimation by making gaze motion an explicit variable rather than an implicit deep feature. It feeds back first-order differences of its own recent predictions as a compact kinematic token, then fuses this with face and eye streams via cross-attention and a causal Transformer decoder. The result is a causal streaming model that cuts mean angular error on Gaze360 by roughly a degree and claims subject-invariant motion via translation-invariant differencing.
Foundation models often rely on spurious correlations, yet generating visual counterfactual explanations is usually slow and entangled. This paper proposes DiDAE, a gradient-free method that wraps a frozen visual foundation model in a conditional diffusion decoder and performs closed-form edits along disentangled dictionary atoms. The approach promises up to 2000× speed-ups over iterative methods while enabling targeted model repair via Counterfactual Knowledge Distillation and unsupervised discovery of shortcut directions.
This paper studies adversarial attacks under a shared global mean-$\ell_1$ budget with per-coordinate magnitude caps. The core idea is to add a centered, clean-gradient importance prior to the attack objective to softly guide perturbation mass toward model-sensitive regions without changing the feasible set. This matters because, under local capacity constraints, simply finding a useful gradient direction is insufficient—attackers must also decide where to spend a budget that cannot be arbitrarily concentrated. Empirically, the mechanism raises attack success by 2.52–17.70 percentage points over matched baselines across ten robust model–dataset configurations.